Pronoun-first sentence starts

Test: Bad Writing Habits

Avg. Score

73.6%

Scenarios

Overall Performance

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

Rank ▲	Model	Score	Avg. Cost	Avg. Time	Stability
1	Claude 3.5 Haiku	97.7%	$0.0035	10.8s	87%
2	Llama 3.1 Nemotron 70B	97.4%	$0.0038	31.7s	83%
3	Llama 3.1 8B	98.4%	$0.0003	1.3m	87%
4	GPT-5.4 Mini (Reasoning)	95.5%	$0.022	28.1s	79%
5	Llama 3.1 70B	96.0%	$0.0015	29.4s	71%
6	GPT-5.4 Mini (Reasoning, Low)	93.3%	$0.015	16.8s	73%
7	GPT-5.4 Nano (Reasoning, Low)	89.7%	$0.0055	20.6s	71%
8	Claude 3 Haiku	90.0%	$0.0025	14.9s	62%
9	GPT-5.4 Mini	91.0%	$0.015	16.8s	64%
10	GPT-5.4 Nano (Reasoning)	87.7%	$0.0061	24.5s	66%
11	Mistral Small 4 (Reasoning)	89.7%	$0.0022	30.2s	63%
12	Mistral Small Creative	87.8%	$0.0007	9.1s	58%
13	Grok 4.1 Fast	90.5%	$0.0018	37.8s	61%
14	Ministral 3B	88.2%	$0.0001	8.1s	56%
15	Ministral 3 14B	86.8%	$0.0007	11.7s	57%
16	DeepSeek V3 (2025-03-24)	91.5%	$0.0014	39.4s	57%
17	Grok 4.20 (Beta)	87.3%	$0.018	15.8s	61%
18	Z.AI GLM 4.5	90.5%	$0.0051	42.1s	58%
19	GPT-5.4 Nano	84.9%	$0.0057	26.3s	61%
20	Z.AI GLM 5 Turbo	89.5%	$0.0081	33.2s	56%
21	Ministral 3 3B	85.9%	$0.0005	11.1s	52%
22	Mistral Small 4	85.4%	$0.0014	18.2s	53%
23	Grok 4.20 (Beta, Reasoning)	88.7%	$0.039	34.0s	67%
24	Mistral Medium 3.1	86.9%	$0.0048	36.5s	54%
25	Mistral Large 3	86.6%	$0.0033	30.3s	51%
26	Claude Haiku 4.5	87.1%	$0.011	21.6s	50%
27	Claude Sonnet 4	90.6%	$0.032	43.7s	58%
28	GPT-5.2	93.6%	$0.056	1.5m	75%
29	Ministral 3 8B	83.7%	$0.0008	19.6s	49%
30	Rocinante 12B	86.6%	$0.0014	38.4s	49%
31	Grok 4 Fast	83.1%	$0.0017	24.1s	50%
32	Claude Sonnet 4.5	91.5%	$0.035	38.1s	55%
33	GPT-5.4 (Reasoning, Low)	93.7%	$0.055	1.4m	71%
34	Qwen 2.5 72B	83.5%	$0.0010	36.7s	51%
35	Mistral Large	85.9%	$0.014	30.9s	51%
36	Mistral Large 2	85.8%	$0.013	29.4s	49%
37	GPT-5.4	92.0%	$0.049	1.4m	69%
38	Ministral 8B	81.3%	$0.0004	10.4s	44%
39	Qwen 3.5 Plus (2026-02-15)	82.2%	$0.0060	31.5s	49%
40	Z.AI GLM 5	86.8%	$0.0084	1.2m	54%
41	Writer: Palmyra X5	83.8%	$0.011	22.0s	43%
42	Claude 3.7 Sonnet	88.2%	$0.042	46.7s	55%
43	Qwen 3 32B	81.1%	$0.0015	54.6s	41%
44	GPT-4o, May 13th (temp=1)	81.1%	$0.033	14.4s	42%
45	GPT-4.1 Mini	78.4%	$0.0027	19.0s	35%
46	Claude Opus 4.5	88.6%	$0.070	53.4s	56%
47	LFM2 24B	77.9%	$0.0002	28.4s	36%
48	Qwen3 235B A22B Instruct 2507	81.3%	$0.0011	59.2s	38%
49	GPT-4o, Aug. 6th (temp=1)	78.4%	$0.018	24.4s	38%
50	GPT-5 Nano	80.9%	$0.0042	1.4m	43%
51	Inception Mercury 2	75.7%	$0.0032	7.0s	28%
52	Hermes 3 405B	77.8%	$0.0032	53.2s	36%
53	MiniMax M2.7	79.4%	$0.0040	1.1m	38%
54	ByteDance Seed 1.6 Flash	75.8%	$0.0013	27.3s	31%
55	MiniMax M2.5	80.5%	$0.0034	1.3m	37%
56	Claude 3.5 Sonnet	82.6%	$0.048	35.5s	37%
57	Cohere Command R+ (Aug. 2024)	77.8%	$0.020	52.5s	36%
58	GPT-4o, Aug. 6th (temp=0)	75.1%	$0.023	22.7s	32%
59	Arcee AI: Trinity Large (Preview)	73.9%	$0.0000	43.6s	30%
60	Hermes 3 70B	76.7%	$0.0010	1.2m	33%
61	Stealth: Aurora Alpha	70.6%	$0.0000	9.8s	24%
62	GPT-5.4 (Reasoning)	91.8%	$0.089	2.6m	64%
63	Gemma 3 12B	70.5%	$0.0004	41.3s	26%
64	Claude Opus 4.6	81.7%	$0.078	1.2m	48%
65	Nemotron 3 Nano	72.6%	$0.0010	1.1m	28%
66	GPT-4o Mini (temp=1)	68.7%	$0.0012	34.8s	23%
67	Claude Opus 4.6 (Reasoning)	83.1%	$0.088	1.4m	49%
68	Gemini 3 Flash (Preview)	63.9%	$0.0078	19.6s	27%
69	DeepSeek V3 (2024-12-26)	70.7%	$0.0021	54.6s	25%
70	Gemini 2.5 Flash	66.9%	$0.0052	10.6s	19%
71	Gemini 2.5 Flash (Reasoning)	67.3%	$0.011	21.5s	24%
72	MoonshotAI: Kimi K2.5	81.0%	$0.019	3.2m	48%
73	Gemini 2.5 Flash Lite	61.6%	$0.0009	9.5s	21%
74	DeepSeek-V2 Chat	68.9%	$0.0021	53.3s	22%
75	Claude Sonnet 4.6	70.1%	$0.031	39.3s	26%
76	Gemini 3 Flash (Preview, Reasoning)	62.3%	$0.012	30.1s	27%
77	GPT-4.1	69.0%	$0.018	44.7s	23%
78	Stealth: Hunter Alpha	66.0%	$0.0000	55.0s	22%
79	Stealth: Healer Alpha	59.7%	$0.0000	23.7s	21%
80	WizardLM 2 8x22b	70.3%	$0.0026	1.8m	29%
81	GPT-5.1	79.1%	$0.054	1.8m	36%
82	Arcee AI: Trinity Mini	57.4%	$0.0003	9.2s	17%
83	GPT-4o, May 13th (temp=0)	66.4%	$0.035	14.1s	19%
84	Gemma 3 27B	61.1%	$0.0006	52.6s	18%
85	o4 Mini	61.2%	$0.015	25.7s	17%
86	Gemini 2.5 Flash Lite (Reasoning)	57.3%	$0.0028	30.8s	18%
87	Nemotron 3 Super	63.8%	$0.0000	1.4m	21%
88	Grok 4	73.8%	$0.048	1.7m	28%
89	Z.AI GLM 4.6	57.8%	$0.0065	51.5s	21%
90	Aion 2.0	63.0%	$0.0064	1.3m	19%
91	GPT-5 Mini	59.6%	$0.0100	57.4s	20%
92	GPT-4o Mini (temp=0)	56.7%	$0.0012	34.8s	14%
93	DeepSeek V3.2	63.0%	$0.0014	1.9m	22%
94	Gemma 3 4B	52.5%	$0.0002	20.0s	11%
95	o4 Mini High	59.5%	$0.025	47.2s	16%
96	GPT-4.1 Nano	46.4%	$0.0007	13.3s	15%
97	Claude Sonnet 4.6 (Reasoning)	65.6%	$0.060	1.2m	22%
98	Qwen 3.5 Flash	48.5%	$0.0025	47.5s	17%
99	Gemini 3.1 Pro (Preview)	75.8%	$0.107	1.8m	36%
100	Mistral NeMO	38.9%	$0.0005	10.1s	16%
101	Z.AI GLM 4.7 Flash	47.9%	$0.0017	1.2m	16%
102	Qwen 3.5 397B A17B	65.1%	$0.014	3.0m	24%
103	Gemini 3 Pro (Preview)	56.8%	$0.055	54.4s	19%
104	Gemini 2.5 Pro	49.0%	$0.036	36.2s	16%
105	Z.AI GLM 4.7	49.8%	$0.010	1.4m	16%
106	Claude Opus 4	87.3%	$0.209	1.4m	44%
107	Gemini 3.1 Flash Lite (Preview)	36.1%	$0.0030	8.4s	11%
108	DeepSeek V3.1	51.7%	$0.0020	1.8m	15%
109	Qwen 3.5 9B	45.8%	$0.0011	1.4m	11%
110	Qwen 3.5 35B	43.3%	$0.018	1.0m	16%
111	Inception Mercury	39.4%	$0.011	17.6s	6%
112	ByteDance Seed 1.6	55.5%	$0.013	2.5m	17%
113	Qwen 3.5 122B	44.3%	$0.025	1.1m	13%
114	Qwen 3.5 27B	46.4%	$0.020	1.6m	15%
115	ByteDance Seed 2.0 Lite	50.7%	$0.012	2.2m	13%
116	GPT-5	51.6%	$0.065	2.8m	17%
117	ByteDance Seed 2.0 Mini	43.3%	$0.0045	4.9m	15%
118	Mistral Small 3.2 24B	42.5%	$0.0069	5.7m	10%
73.64%

Individual Scenarios

Detailed Writing Rules

▼

Fantasy: entering an ancient ruin

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.6 (Reasoning)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
Claude Sonnet 4.6 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.1	100	100	100	100	100	100.0%
Claude Opus 4.6	100	100	100	100	100	100.0%
Qwen 3.5 397B A17B	100	100	100	100	100	100.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Z.AI GLM 5	100	100	100	100	100	100.0%
MoonshotAI: Kimi K2.5	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
o4 Mini High	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
MiniMax M2.7	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
MiniMax M2.5	100	100	100	100	100	100.0%
GPT-4.1	100	100	100	100	100	100.0%
o4 Mini	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
DeepSeek-V2 Chat	100	100	100	100	100	100.0%
Nemotron 3 Super	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Inception Mercury 2	100	100	100	100	100	100.0%
GPT-4o, May 13th (temp=1)	100	100	100	100	100	100.0%
Stealth: Aurora Alpha	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
DeepSeek V3 (2024-12-26)	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	100	100.0%
GPT-5 Nano	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=0)	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
Mistral Small 4 (Reasoning)	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Gemini 2.5 Flash Lite	100	100	100	100	100	100.0%
Mistral Large	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
Writer: Palmyra X5	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	100	100	100.0%
GPT-4o Mini (temp=1)	100	100	100	100	100	100.0%
Gemma 3 12B	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
GPT-4o Mini (temp=0)	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Nemotron 3 Nano	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Qwen 2.5 72B	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
GPT-5.4 Nano	100	100	100	100	100	100.0%
Mistral Small Creative	100	100	100	100	100	100.0%
Hermes 3 70B	100	100	100	100	100	100.0%
Ministral 3 14B	100	100	100	100	100	100.0%
Ministral 3 8B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Ministral 3B	100	100	100	100	100	100.0%
LFM2 24B	100	100	100	100	100	100.0%
Rocinante 12B	100	100	100	100	100	100.0%
Ministral 3 3B	100	100	100	100	99	99.8%
Claude 3.5 Sonnet	100	100	100	100	99	99.8%
Claude 3 Haiku	100	100	100	100	99	99.7%
Claude Sonnet 4.6	100	100	100	100	98	99.6%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	98	99.5%
GPT-5.4 Nano (Reasoning)	100	100	100	100	97	99.5%
Gemini 2.5 Flash	100	100	100	100	96	99.2%
GPT-4.1 Mini	100	100	100	100	95	98.9%
Aion 2.0	100	100	100	99	95	98.8%
Gemini 2.5 Pro	100	100	100	97	96	98.6%
Grok 4 Fast	100	100	100	100	92	98.4%
DeepSeek V3.1	100	100	97	97	97	98.2%
Gemini 2.5 Flash Lite (Reasoning)	100	100	100	100	89	97.8%
Arcee AI: Trinity Mini	100	100	100	100	89	97.8%
Gemma 3 4B	100	100	100	100	88	97.7%
GPT-4.1 Nano	100	100	100	95	93	97.5%
Qwen 3 32B	100	100	100	100	88	97.5%
Hermes 3 405B	100	100	100	100	84	96.9%
Stealth: Healer Alpha	100	100	100	94	87	96.2%
ByteDance Seed 1.6 Flash	100	100	100	100	72	94.5%
Cohere Command R+ (Aug. 2024)	100	100	100	93	79	94.4%
Ministral 8B	100	100	100	100	72	94.4%
GPT-5.2	100	100	100	100	70	94.1%
DeepSeek V3.2	100	100	95	95	80	93.9%
Gemini 3 Flash (Preview, Reasoning)	100	100	100	100	67	93.4%
Gemini 3 Flash (Preview)	100	100	100	100	66	93.3%
WizardLM 2 8x22b	100	100	100	100	63	92.7%
GPT-5	100	100	100	100	56	91.2%
Stealth: Hunter Alpha	100	100	100	84	71	90.9%
Gemma 3 27B	100	100	100	87	68	90.9%
Qwen 3.5 122B	100	100	90	89	73	90.6%
Z.AI GLM 4.6	100	100	100	97	54	90.1%
Gemini 2.5 Flash (Reasoning)	100	100	100	100	46	89.2%
ByteDance Seed 1.6	100	100	93	78	75	89.2%
GPT-5 Mini	100	100	97	75	71	88.8%
Arcee AI: Trinity Large (Preview)	100	100	100	86	54	87.9%
ByteDance Seed 2.0 Lite	100	100	100	100	27	85.5%
GPT-4o, May 13th (temp=0)	100	100	100	83	18	80.1%
Qwen 3.5 9B	100	100	85	59	55	79.9%
Gemini 3.1 Flash Lite (Preview)	100	100	100	46	43	77.7%
Qwen 3.5 Flash	100	100	93	45	43	76.2%
Inception Mercury	100	100	75	55	48	75.3%
Qwen 3.5 35B	100	96	72	66	40	74.7%
Z.AI GLM 4.7 Flash	100	100	81	47	24	70.4%
Gemini 3.1 Pro (Preview)	85	81	72	55	51	68.7%
Qwen 3.5 27B	98	84	74	49	23	65.7%
Gemini 3 Pro (Preview)	100	98	51	44	26	63.8%
Mistral Small 3.2 24B	93	85	66	22	20	57.1%
Mistral NeMO	100	100	20	11	6	47.5%
Z.AI GLM 4.7	100	42	33	28	26	45.8%
ByteDance Seed 2.0 Mini	79	42	34	33	30	43.6%

▼

Horror: alone in an eerie place at night

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.6 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
MiniMax M2.7	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
MiniMax M2.5	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
GPT-4.1 Mini	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
GPT-4o Mini (temp=1)	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Mistral Small Creative	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Stealth: Hunter Alpha	100	100	100	100	99	99.8%
Writer: Palmyra X5	100	100	100	100	99	99.7%
Z.AI GLM 5	100	100	100	100	98	99.7%
Mistral Large	100	100	100	100	97	99.3%
MoonshotAI: Kimi K2.5	100	100	100	100	97	99.3%
Z.AI GLM 5 Turbo	100	100	100	100	95	98.9%
Claude Opus 4.5	100	100	100	100	94	98.9%
Claude 3.7 Sonnet	100	100	100	100	94	98.9%
Claude Opus 4.6	100	100	100	100	94	98.7%
ByteDance Seed 1.6 Flash	100	100	100	99	94	98.6%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	91	98.2%
Ministral 3 3B	100	100	100	100	90	98.1%
Qwen 3 32B	100	100	100	100	89	97.9%
Grok 4.1 Fast	100	100	100	98	92	97.9%
Hermes 3 405B	100	100	100	100	89	97.8%
Ministral 3 14B	100	100	100	96	93	97.7%
GPT-5.2	100	100	100	98	87	97.2%
Inception Mercury 2	100	100	100	97	88	97.1%
Ministral 8B	100	100	100	97	88	97.0%
Llama 3.1 8B	100	100	100	100	84	96.9%
Ministral 3B	100	100	100	97	87	96.7%
Stealth: Aurora Alpha	100	100	100	100	83	96.5%
Grok 4 Fast	100	100	98	96	88	96.5%
Mistral Small 4 (Reasoning)	100	100	100	92	89	96.1%
GPT-4o, May 13th (temp=1)	100	100	97	94	87	95.7%
LFM2 24B	100	100	98	96	81	95.1%
Qwen 3.5 Plus (2026-02-15)	100	100	100	88	84	94.5%
Mistral Small 4	100	100	100	88	80	93.5%
Qwen 3.5 397B A17B	99	98	93	91	87	93.4%
WizardLM 2 8x22b	100	100	95	90	83	93.4%
GPT-5.4 Nano	100	100	95	94	76	93.1%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	88	76	92.7%
GPT-4o, Aug. 6th (temp=1)	100	100	100	88	76	92.7%
Ministral 3 8B	100	100	97	85	81	92.6%
Grok 4.20 (Beta)	100	100	100	90	72	92.5%
Llama 3.1 70B	100	100	100	94	67	92.2%
GPT-4.1	100	100	100	87	70	91.3%
DeepSeek V3 (2024-12-26)	100	100	100	80	68	89.4%
Cohere Command R+ (Aug. 2024)	100	100	96	96	55	89.3%
Aion 2.0	100	100	86	78	77	88.1%
Grok 4	100	100	92	77	69	87.6%
GPT-5 Nano	100	100	88	82	66	87.2%
GPT-5.1	100	84	84	84	83	87.0%
Gemini 2.5 Flash	100	99	94	74	67	87.0%
Gemma 3 27B	100	100	84	76	70	86.0%
Rocinante 12B	100	100	100	100	29	85.9%
Claude Sonnet 4.6	100	100	80	79	65	84.6%
Gemini 3.1 Pro (Preview)	100	98	92	75	55	84.0%
Gemini 3 Flash (Preview)	91	85	85	83	75	83.8%
Qwen 2.5 72B	100	93	85	80	59	83.5%
Grok 4.20 (Beta, Reasoning)	91	86	84	79	76	83.2%
DeepSeek V3.2	100	90	78	75	73	83.1%
DeepSeek-V2 Chat	100	100	87	64	62	82.6%
Hermes 3 70B	100	100	98	92	17	81.3%
Nemotron 3 Super	100	89	85	67	61	80.4%
GPT-5.4 Nano (Reasoning)	98	97	85	61	58	80.1%
Arcee AI: Trinity Mini	100	100	80	60	60	80.1%
o4 Mini	100	100	88	77	21	77.2%
Gemini 2.5 Flash Lite (Reasoning)	95	85	80	77	46	76.5%
o4 Mini High	100	75	73	70	65	76.3%
Stealth: Healer Alpha	100	86	81	57	56	76.0%
GPT-4o Mini (temp=0)	92	74	73	67	66	74.3%
Claude Sonnet 4.6 (Reasoning)	100	80	79	59	52	74.0%
GPT-5 Mini	100	80	71	65	50	73.3%
Gemini 2.5 Flash (Reasoning)	96	83	79	71	35	72.7%
Z.AI GLM 4.6	93	79	68	67	54	72.5%
Arcee AI: Trinity Large (Preview)	100	100	80	57	24	72.4%
Gemma 3 4B	98	85	64	54	54	70.9%
Gemma 3 12B	98	95	73	45	43	70.8%
DeepSeek V3.1	100	75	70	66	41	70.5%
Inception Mercury	100	99	90	63	0	70.3%
GPT-4o, Aug. 6th (temp=0)	88	87	62	59	51	69.5%
Gemini 2.5 Pro	88	72	68	59	50	67.5%
GPT-4.1 Nano	100	73	61	59	44	67.4%
Gemini 3 Flash (Preview, Reasoning)	100	88	72	54	0	62.8%
Nemotron 3 Nano	100	82	75	51	5	62.5%
GPT-5	76	74	62	51	50	62.4%
ByteDance Seed 1.6	84	71	60	57	38	62.1%
Gemini 2.5 Flash Lite	89	60	60	51	44	60.6%
Z.AI GLM 4.7 Flash	75	71	57	55	14	54.4%
Gemini 3 Pro (Preview)	70	64	55	35	32	51.2%
GPT-4o, May 13th (temp=0)	95	87	45	26	0	50.7%
Qwen 3.5 122B	78	49	47	39	32	49.1%
Qwen 3.5 9B	100	100	45	0	0	49.0%
Qwen 3.5 Flash	61	59	49	48	23	47.9%
Qwen 3.5 35B	79	74	61	25	0	47.8%
Mistral NeMO	72	57	54	44	0	45.6%
Qwen 3.5 27B	64	50	43	38	22	43.4%
Z.AI GLM 4.7	68	55	46	41	0	42.0%
ByteDance Seed 2.0 Mini	40	23	20	13	6	20.3%
Mistral Small 3.2 24B	54	30	16	0	0	20.0%
ByteDance Seed 2.0 Lite	44	28	14	5	0	18.3%
Gemini 3.1 Flash Lite (Preview)	36	20	15	11	0	16.4%

▼

Literary fiction: old friends reunite

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
Claude Sonnet 4.6	100	100	100	100	100	100.0%
GPT-5.2	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
Inception Mercury 2	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	98	99.6%
Llama 3.1 70B	100	100	100	100	98	99.5%
Llama 3.1 8B	100	100	100	100	96	99.2%
GPT-5.4 (Reasoning, Low)	100	100	100	100	94	98.8%
Mistral Large 3	100	100	100	100	82	96.5%
Gemini 2.5 Flash (Reasoning)	100	100	100	100	82	96.4%
GPT-5.4 Nano (Reasoning)	100	100	100	100	82	96.3%
Claude Sonnet 4	100	100	100	100	76	95.2%
GPT-5.4 Nano	100	100	96	90	87	94.6%
Stealth: Aurora Alpha	100	100	100	90	83	94.5%
GPT-5.4 Nano (Reasoning, Low)	100	98	98	90	83	94.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	68	93.6%
Z.AI GLM 4.5	100	100	100	100	65	93.1%
Ministral 3 14B	100	100	97	84	80	92.3%
Mistral Small Creative	100	96	96	85	83	92.0%
Claude Sonnet 4.5	100	100	100	100	58	91.5%
Rocinante 12B	100	100	100	100	57	91.4%
Qwen 2.5 72B	100	100	100	100	57	91.3%
Writer: Palmyra X5	100	100	100	97	59	91.2%
GPT-4o, Aug. 6th (temp=0)	100	100	100	99	56	91.1%
Claude Opus 4.6	100	100	100	100	49	89.8%
DeepSeek V3 (2025-03-24)	100	100	96	93	60	89.7%
GPT-5.4	100	100	98	86	65	89.7%
Claude Sonnet 4.6 (Reasoning)	100	100	100	92	55	89.4%
Mistral Small 4 (Reasoning)	100	100	100	82	63	89.1%
Ministral 3 3B	100	100	92	78	73	88.7%
Grok 4.20 (Beta)	100	100	88	84	72	88.7%
Ministral 3B	100	100	100	100	42	88.4%
MiniMax M2.5	100	100	96	88	56	87.9%
MoonshotAI: Kimi K2.5	100	100	100	69	64	86.6%
Grok 4.20 (Beta, Reasoning)	100	100	95	78	52	85.1%
Claude Haiku 4.5	100	100	100	100	25	84.9%
Claude 3 Haiku	96	94	84	84	66	84.8%
Claude Opus 4.5	100	100	87	72	63	84.5%
Z.AI GLM 5 Turbo	100	100	100	80	39	83.9%
Mistral Small 4	100	100	100	84	33	83.5%
Ministral 8B	100	100	100	85	28	82.6%
Mistral Medium 3.1	100	100	100	59	54	82.6%
GPT-5.4 Mini	100	100	72	70	65	81.2%
ByteDance Seed 1.6 Flash	100	100	100	88	13	80.2%
GPT-4o, May 13th (temp=0)	100	100	100	100	0	80.0%
Claude Opus 4	100	100	100	100	0	79.9%
Grok 4.1 Fast	100	100	88	65	46	79.8%
o4 Mini High	100	100	100	96	0	79.2%
Ministral 3 8B	100	93	79	74	49	79.1%
Qwen 3 32B	100	98	95	82	13	77.5%
GPT-4.1 Mini	98	85	82	60	56	76.2%
Nemotron 3 Nano	100	85	70	62	61	75.4%
ByteDance Seed 1.6	100	100	66	60	46	74.4%
Qwen 3.5 Plus (2026-02-15)	100	84	79	68	38	73.8%
Mistral Large	100	100	67	52	50	73.7%
MiniMax M2.7	100	100	100	46	22	73.6%
Grok 4 Fast	100	100	64	53	46	72.7%
Claude Opus 4.6 (Reasoning)	100	79	65	63	52	71.8%
Claude 3.7 Sonnet	100	83	67	52	49	70.0%
GPT-5 Nano	98	82	63	63	35	68.3%
WizardLM 2 8x22b	100	94	40	39	37	62.1%
Mistral Small 3.2 24B	100	100	61	22	21	60.7%
Z.AI GLM 5	100	66	53	46	38	60.5%
GPT-4.1	100	100	100	1	0	60.3%
GPT-4o, Aug. 6th (temp=1)	100	100	45	39	17	60.2%
Mistral Large 2	100	93	44	36	29	60.2%
GPT-5.1	100	100	100	0	0	60.0%
o4 Mini	100	100	97	1	0	59.6%
GPT-4o Mini (temp=1)	100	100	74	17	6	59.4%
GPT-4o, May 13th (temp=1)	100	89	68	38	0	58.9%
Qwen 3.5 397B A17B	89	62	53	43	38	57.0%
ByteDance Seed 2.0 Mini	100	100	81	0	0	56.2%
Cohere Command R+ (Aug. 2024)	100	96	49	30	0	55.2%
DeepSeek-V2 Chat	100	85	65	26	0	55.2%
ByteDance Seed 2.0 Lite	100	84	64	17	0	53.1%
DeepSeek V3 (2024-12-26)	100	100	61	0	0	52.3%
DeepSeek V3.2	100	88	64	0	0	50.5%
Hermes 3 405B	67	64	50	35	32	49.6%
Claude 3.5 Sonnet	76	64	44	44	16	48.9%
Qwen 3.5 9B	96	84	48	16	0	48.8%
Gemini 3.1 Pro (Preview)	100	55	49	37	0	48.2%
Z.AI GLM 4.6	100	77	20	20	10	45.4%
Hermes 3 70B	99	58	38	31	0	45.1%
Gemini 3 Flash (Preview)	83	60	37	36	0	43.1%
Gemini 3 Flash (Preview, Reasoning)	79	75	30	15	11	42.0%
Z.AI GLM 4.7 Flash	81	64	52	12	0	41.9%
LFM2 24B	100	49	28	17	15	41.8%
Gemini 2.5 Flash	100	100	8	0	0	41.6%
Grok 4	100	47	46	4	0	39.5%
GPT-5 Mini	100	57	30	5	1	38.5%
Mistral NeMO	94	52	33	2	0	36.3%
Qwen 3.5 122B	52	45	38	35	7	35.4%
GPT-5	73	42	28	20	8	34.2%
Qwen 3.5 27B	100	37	30	0	0	33.4%
Gemini 2.5 Flash Lite (Reasoning)	100	25	20	18	3	33.2%
Gemini 2.5 Flash Lite	95	42	15	8	0	32.1%
Qwen 3.5 Flash	70	26	18	15	14	28.8%
Stealth: Hunter Alpha	77	35	12	9	0	26.7%
Nemotron 3 Super	69	50	3	0	0	24.4%
Qwen 3.5 35B	61	57	2	0	0	23.9%
Aion 2.0	98	15	3	0	0	23.4%
Arcee AI: Trinity Large (Preview)	100	5	0	0	0	21.1%
Stealth: Healer Alpha	36	23	22	12	4	19.5%
Gemini 2.5 Pro	69	11	0	0	0	16.0%
Gemini 3 Pro (Preview)	67	0	0	0	0	13.5%
Gemma 3 12B	23	15	5	0	0	8.6%
Gemma 3 27B	24	9	8	0	0	8.2%
Arcee AI: Trinity Mini	17	7	7	0	0	6.1%
GPT-4o Mini (temp=0)	24	0	0	0	0	4.9%
Gemini 3.1 Flash Lite (Preview)	20	3	0	0	0	4.6%
DeepSeek V3.1	14	0	0	0	0	2.9%
Inception Mercury	7	7	0	0	0	2.8%
Z.AI GLM 4.7	7	0	0	0	0	1.4%
GPT-4.1 Nano	2	2	0	0	0	0.9%
Gemma 3 4B	0	0	0	0	0	0.0%

▼

Mystery: examining a crime scene

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.6 (Reasoning)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Z.AI GLM 5	100	100	100	100	100	100.0%
MoonshotAI: Kimi K2.5	100	100	100	100	100	100.0%
GPT-5.2	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
GPT-4o, May 13th (temp=0)	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
DeepSeek-V2 Chat	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Inception Mercury 2	100	100	100	100	100	100.0%
Stealth: Aurora Alpha	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
DeepSeek V3 (2024-12-26)	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4.1 Mini	100	100	100	100	100	100.0%
GPT-5 Nano	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Small 4 (Reasoning)	100	100	100	100	100	100.0%
Qwen 3 32B	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning)	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
Writer: Palmyra X5	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Nemotron 3 Nano	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
GPT-5.4 Nano	100	100	100	100	100	100.0%
ByteDance Seed 1.6 Flash	100	100	100	100	100	100.0%
Ministral 3 14B	100	100	100	100	100	100.0%
Ministral 3 8B	100	100	100	100	100	100.0%
Ministral 3 3B	100	100	100	100	100	100.0%
Ministral 8B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Ministral 3B	100	100	100	100	100	100.0%
LFM2 24B	100	100	100	100	100	100.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	98	99.6%
Claude 3 Haiku	100	100	100	100	97	99.4%
MiniMax M2.7	100	100	100	100	93	98.6%
GPT-5 Mini	100	100	100	99	92	98.2%
GPT-5.1	100	100	100	100	91	98.1%
Nemotron 3 Super	100	100	100	100	89	97.8%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	89	97.8%
Stealth: Hunter Alpha	100	100	100	100	88	97.6%
GPT-4o, May 13th (temp=1)	100	100	100	100	87	97.3%
GPT-4o, Aug. 6th (temp=0)	100	100	100	100	87	97.3%
GPT-5.4 Mini (Reasoning)	100	100	100	100	83	96.6%
Aion 2.0	100	100	100	92	90	96.5%
Claude Opus 4.6	100	100	100	100	80	95.9%
Gemini 2.5 Flash Lite	100	100	100	91	84	94.9%
Mistral Small 3.2 24B	100	100	100	100	71	94.3%
MiniMax M2.5	100	100	100	100	71	94.3%
Grok 4 Fast	100	100	100	90	81	94.3%
DeepSeek V3.1	100	100	100	100	71	94.1%
Hermes 3 70B	100	100	100	94	75	93.6%
Qwen 3.5 397B A17B	100	100	96	92	77	93.0%
GPT-4o Mini (temp=0)	100	100	100	85	78	92.5%
Qwen 2.5 72B	100	100	100	80	78	91.7%
Claude Sonnet 4.6	100	100	96	92	70	91.7%
o4 Mini High	100	100	100	100	57	91.4%
Claude Sonnet 4.6 (Reasoning)	100	100	100	100	56	91.2%
Gemini 2.5 Flash	100	100	100	82	75	91.2%
WizardLM 2 8x22b	100	98	96	92	69	91.2%
Rocinante 12B	100	100	99	90	60	89.7%
Gemini 2.5 Flash (Reasoning)	100	97	88	81	77	88.6%
ByteDance Seed 2.0 Lite	100	97	91	83	73	88.6%
o4 Mini	100	100	100	73	60	86.7%
Gemini 3 Pro (Preview)	100	88	83	82	78	86.2%
DeepSeek V3.2	98	93	88	76	75	86.1%
Cohere Command R+ (Aug. 2024)	100	100	96	78	56	85.9%
Mistral Small Creative	100	100	88	76	62	85.1%
Gemini 3 Flash (Preview)	100	99	79	76	65	83.8%
Gemma 3 12B	100	100	99	60	51	82.0%
Hermes 3 405B	100	100	98	70	37	80.9%
GPT-5	100	89	80	77	59	80.9%
Z.AI GLM 4.7	100	91	81	74	55	80.2%
ByteDance Seed 1.6	100	100	80	71	37	77.6%
Z.AI GLM 4.7 Flash	90	88	84	84	37	76.5%
GPT-4o Mini (temp=1)	91	89	71	66	57	74.7%
Qwen 3.5 122B	100	92	91	67	22	74.5%
Mistral Large	100	84	68	61	55	73.8%
Gemini 3 Flash (Preview, Reasoning)	100	82	78	75	33	73.4%
Mistral Large 2	100	100	100	45	20	73.0%
Qwen 3.5 Flash	100	100	79	60	23	72.3%
Mistral NeMO	93	72	71	64	56	71.2%
Arcee AI: Trinity Mini	100	100	70	44	41	71.0%
Z.AI GLM 4.6	96	87	82	45	44	70.8%
Stealth: Healer Alpha	92	84	69	59	47	70.0%
GPT-4.1	88	80	72	59	49	69.6%
Arcee AI: Trinity Large (Preview)	89	71	67	60	51	67.5%
ByteDance Seed 2.0 Mini	90	80	58	52	45	65.1%
Gemini 3.1 Pro (Preview)	100	100	42	41	40	64.6%
Gemma 3 4B	100	72	63	45	42	64.3%
Gemini 2.5 Flash Lite (Reasoning)	100	81	74	31	20	61.3%
Mistral Large 3	88	68	65	47	36	60.8%
Qwen 3.5 9B	100	100	100	0	0	60.0%
Gemini 2.5 Pro	80	68	55	49	47	59.7%
Qwen 3.5 35B	100	64	57	41	34	59.1%
Inception Mercury	100	100	46	46	0	58.5%
Gemma 3 27B	100	78	49	43	12	56.4%
Gemini 3.1 Flash Lite (Preview)	100	90	42	29	15	55.3%
Qwen 3.5 27B	78	62	57	47	0	48.7%
GPT-4.1 Nano	65	50	44	33	26	43.7%

▼

Romance: separated couple reunites

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude 3.5 Haiku	100	100	100	100	94	98.7%
Mistral Small Creative	100	100	100	98	96	98.7%
Llama 3.1 70B	100	100	100	100	92	98.3%
GPT-5.4 Nano (Reasoning)	100	100	100	100	91	98.2%
GPT-5.4 Nano	100	100	99	98	91	97.6%
Llama 3.1 8B	100	100	100	100	75	94.9%
Grok 4.1 Fast	100	100	100	99	71	94.0%
Claude Sonnet 4.5	100	100	100	98	65	92.6%
Claude 3 Haiku	100	100	100	81	79	92.0%
Z.AI GLM 4.5	100	100	100	91	67	91.6%
DeepSeek V3 (2025-03-24)	100	100	97	82	79	91.6%
GPT-5.4	100	100	100	84	73	91.5%
GPT-5.2	100	100	95	86	67	89.5%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	92	48	87.9%
GPT-5.4 (Reasoning, Low)	100	96	91	77	72	87.2%
Llama 3.1 Nemotron 70B	100	100	99	72	51	84.3%
Claude Sonnet 4	100	100	100	74	46	84.0%
Mistral Small 4 (Reasoning)	100	100	100	66	53	83.8%
GPT-5.4 Mini (Reasoning)	100	85	84	79	59	81.2%
GPT-5 Nano	100	100	95	73	34	80.4%
GPT-5.4 Mini	100	90	85	67	58	79.9%
Inception Mercury 2	96	93	77	67	64	79.3%
Ministral 3 3B	100	100	100	96	0	79.1%
Ministral 3 8B	100	100	68	64	60	78.4%
Stealth: Aurora Alpha	100	100	78	60	51	77.9%
Ministral 3 14B	100	99	93	54	41	77.5%
Grok 4.20 (Beta, Reasoning)	100	93	69	62	57	76.0%
Rocinante 12B	100	100	96	73	3	74.6%
Cohere Command R+ (Aug. 2024)	100	100	90	52	30	74.3%
Ministral 8B	100	94	87	51	35	73.5%
GPT-5.4 Mini (Reasoning, Low)	100	88	81	76	17	72.4%
Mistral Large 3	100	100	91	67	0	71.5%
Claude Haiku 4.5	100	89	85	46	35	71.0%
Writer: Palmyra X5	100	76	70	56	52	70.8%
Qwen 3.5 Plus (2026-02-15)	98	87	67	53	49	70.7%
Z.AI GLM 5 Turbo	93	87	65	52	42	67.8%
Claude Opus 4.6	100	65	62	61	46	66.8%
Mistral Large	100	79	53	49	38	63.7%
Mistral Small 4	100	85	52	45	33	63.0%
Mistral Medium 3.1	87	70	62	56	38	62.6%
GPT-4o, May 13th (temp=0)	100	76	70	62	0	61.6%
Qwen 2.5 72B	90	81	69	67	0	61.4%
ByteDance Seed 1.6	100	100	71	24	10	61.0%
Ministral 3B	100	98	44	43	20	60.9%
Grok 4.20 (Beta)	84	77	49	47	45	60.5%
Z.AI GLM 5	100	92	83	17	9	60.3%
GPT-5.4 (Reasoning)	78	70	64	61	23	59.4%
Hermes 3 70B	100	90	51	47	0	57.5%
Claude Opus 4.6 (Reasoning)	84	83	44	40	34	57.3%
Claude 3.5 Sonnet	100	99	55	16	16	57.3%
Claude 3.7 Sonnet	96	85	51	44	11	57.3%
Nemotron 3 Nano	100	100	75	6	1	56.4%
Stealth: Hunter Alpha	100	79	64	28	0	54.3%
Qwen 3 32B	87	79	61	44	0	54.1%
Qwen3 235B A22B Instruct 2507	85	65	52	39	23	52.6%
MiniMax M2.5	75	64	60	51	12	52.4%
GPT-4.1 Mini	100	77	35	24	20	51.1%
Claude Opus 4	100	73	66	15	0	50.9%
MoonshotAI: Kimi K2.5	75	73	47	32	20	49.4%
GPT-4o, Aug. 6th (temp=1)	100	81	46	20	0	49.3%
Claude Opus 4.5	85	58	40	33	29	49.0%
GPT-4o, May 13th (temp=1)	100	66	45	20	9	47.8%
Grok 4 Fast	63	47	44	42	34	46.1%
Arcee AI: Trinity Mini	84	62	40	39	0	45.1%
Claude Sonnet 4.6	77	47	47	37	17	45.0%
Arcee AI: Trinity Large (Preview)	83	62	50	27	0	44.4%
Mistral Small 3.2 24B	100	48	38	18	0	41.0%
DeepSeek-V2 Chat	100	100	0	0	0	40.0%
GPT-4o, Aug. 6th (temp=0)	84	48	41	25	0	39.7%
Grok 4	66	42	35	35	18	39.1%
LFM2 24B	67	63	45	14	0	37.6%
Qwen 3.5 27B	71	43	38	34	0	37.3%
Mistral Large 2	78	47	26	24	0	34.9%
Mistral NeMO	67	57	41	5	2	34.2%
WizardLM 2 8x22b	65	32	31	25	0	30.5%
Stealth: Healer Alpha	94	43	10	2	0	29.8%
Gemini 2.5 Flash Lite	46	39	34	27	0	29.2%
Aion 2.0	54	46	43	3	0	29.2%
Z.AI GLM 4.6	53	35	30	26	0	28.6%
o4 Mini	63	52	27	0	0	28.4%
ByteDance Seed 1.6 Flash	50	44	30	13	0	27.3%
Qwen 3.5 9B	70	58	0	0	0	25.4%
Qwen 3.5 Flash	60	49	16	2	0	25.3%
DeepSeek V3.2	53	43	26	3	0	24.8%
Qwen 3.5 397B A17B	36	33	26	19	10	24.5%
Qwen 3.5 35B	54	30	25	13	0	24.4%
GPT-5 Mini	37	37	33	15	0	24.3%
GPT-5.1	64	22	20	14	1	23.9%
MiniMax M2.7	62	54	2	0	0	23.7%
Qwen 3.5 122B	70	31	4	0	0	20.9%
Z.AI GLM 4.7 Flash	42	36	20	0	0	19.7%
Inception Mercury	70	27	0	0	0	19.4%
Gemini 2.5 Flash (Reasoning)	43	42	6	0	0	18.2%
Claude Sonnet 4.6 (Reasoning)	24	20	15	11	10	15.9%
Gemini 3.1 Pro (Preview)	54	15	0	0	0	13.8%
Gemini 3 Flash (Preview)	68	0	0	0	0	13.5%
Nemotron 3 Super	41	20	0	0	0	12.3%
Gemma 3 12B	49	13	0	0	0	12.2%
GPT-4o Mini (temp=1)	33	24	1	0	0	11.4%
Gemini 2.5 Flash	28	26	2	0	0	11.3%
DeepSeek V3.1	25	22	0	0	0	9.4%
GPT-5	24	12	8	0	0	8.5%
Z.AI GLM 4.7	24	16	0	0	0	8.0%
GPT-4.1	32	4	2	0	0	7.6%
Hermes 3 405B	20	17	0	0	0	7.3%
Gemini 2.5 Flash Lite (Reasoning)	36	0	0	0	0	7.1%
Gemma 3 4B	28	1	0	0	0	5.9%
GPT-4.1 Nano	23	3	0	0	0	5.3%
Gemini 2.5 Pro	20	6	0	0	0	5.2%
Gemma 3 27B	11	6	0	0	0	3.4%
Gemini 3 Pro (Preview)	7	0	0	0	0	1.3%
GPT-4o Mini (temp=0)	6	0	0	0	0	1.1%
DeepSeek V3 (2024-12-26)	5	0	0	0	0	1.0%
o4 Mini High	3	0	0	0	0	0.6%
Gemini 3 Flash (Preview, Reasoning)	0	0	0	0	0	0.0%
ByteDance Seed 2.0 Mini	0	0	0	0	0	0.0%
Gemini 3.1 Flash Lite (Preview)	0	0	0	0	0	0.0%
ByteDance Seed 2.0 Lite	0	0	0	0	0	0.0%

▼

Thriller: chase through city streets

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
MiniMax M2.7	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
GPT-4.1	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Inception Mercury 2	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4.1 Mini	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
Writer: Palmyra X5	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Nemotron 3 Nano	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	99.9%
MoonshotAI: Kimi K2.5	100	100	100	100	99	99.8%
Claude Haiku 4.5	100	100	100	100	99	99.8%
Mistral Medium 3.1	100	100	100	100	98	99.6%
Qwen 3 32B	100	100	100	100	98	99.5%
Grok 4	100	100	100	99	98	99.5%
Rocinante 12B	100	100	100	100	97	99.5%
Z.AI GLM 4.5	100	100	100	99	97	99.3%
Grok 4.20 (Beta)	100	100	100	100	96	99.2%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	95	99.0%
Gemma 3 4B	100	100	100	100	94	98.8%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	93	98.5%
Claude Opus 4.6	100	100	100	97	96	98.5%
Mistral Small Creative	100	100	100	100	92	98.4%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	92	98.3%
Stealth: Aurora Alpha	100	100	100	100	91	98.2%
Ministral 8B	100	100	100	100	90	98.0%
MiniMax M2.5	100	100	100	100	89	97.7%
GPT-5.2	100	100	100	95	94	97.7%
Z.AI GLM 5	100	100	100	100	87	97.3%
Hermes 3 405B	100	100	100	100	85	97.0%
Ministral 3 3B	100	100	100	100	83	96.7%
Ministral 3B	100	100	100	100	82	96.3%
Claude Opus 4.6 (Reasoning)	100	100	100	100	81	96.3%
ByteDance Seed 1.6 Flash	100	100	100	95	84	95.8%
Mistral Small 4 (Reasoning)	100	100	100	100	76	95.2%
Aion 2.0	100	100	100	89	85	94.7%
Ministral 3 8B	100	100	100	100	72	94.5%
Ministral 3 14B	100	100	100	100	72	94.4%
GPT-4o, Aug. 6th (temp=1)	100	100	100	93	77	93.9%
Mistral Large	100	100	100	100	70	93.9%
GPT-5.4 Nano (Reasoning)	100	97	92	90	90	93.8%
LFM2 24B	100	100	100	93	76	93.8%
Claude Sonnet 4.6 (Reasoning)	100	100	100	94	75	93.8%
Llama 3.1 8B	100	100	100	100	68	93.7%
Z.AI GLM 5 Turbo	100	99	92	90	85	93.3%
Stealth: Hunter Alpha	100	100	100	80	80	91.9%
WizardLM 2 8x22b	100	100	97	93	68	91.5%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	96	58	90.8%
Gemini 2.5 Flash (Reasoning)	100	100	100	86	65	90.1%
Qwen 3.5 397B A17B	100	97	93	91	68	89.9%
Grok 4 Fast	100	100	100	81	62	88.5%
GPT-4o, Aug. 6th (temp=0)	100	100	100	76	64	87.9%
Gemini 2.5 Flash Lite (Reasoning)	100	100	92	84	62	87.6%
Qwen 2.5 72B	100	100	88	80	70	87.6%
GPT-4o, May 13th (temp=1)	100	91	90	74	73	85.5%
Gemini 3 Flash (Preview)	98	97	88	74	70	85.3%
Gemini 3.1 Pro (Preview)	94	93	86	81	64	83.4%
Gemini 2.5 Flash	100	87	83	82	64	83.3%
Nemotron 3 Super	100	89	87	83	54	82.6%
DeepSeek V3.2	100	96	85	68	64	82.5%
Cohere Command R+ (Aug. 2024)	100	100	100	68	43	82.2%
GPT-4o, May 13th (temp=0)	100	92	86	84	49	82.1%
Z.AI GLM 4.6	100	97	85	65	61	81.6%
GPT-4o Mini (temp=1)	100	89	79	71	67	81.2%
GPT-5.4 Nano	93	87	81	80	64	80.9%
Gemini 2.5 Flash Lite	90	90	85	78	62	80.8%
ByteDance Seed 1.6	100	87	83	81	52	80.5%
Hermes 3 70B	100	87	87	75	44	78.4%
Gemma 3 12B	100	100	72	61	59	78.3%
GPT-5 Nano	92	84	76	73	64	77.5%
Arcee AI: Trinity Mini	100	100	87	79	20	77.1%
DeepSeek V3 (2024-12-26)	100	100	100	54	29	76.7%
Claude Sonnet 4.6	100	94	71	62	50	75.4%
Gemini 3 Flash (Preview, Reasoning)	94	91	82	62	47	75.3%
GPT-5.1	80	79	77	64	62	72.4%
DeepSeek-V2 Chat	100	100	62	48	46	71.3%
o4 Mini	93	84	81	48	47	70.7%
GPT-4o Mini (temp=0)	84	77	70	57	56	68.6%
DeepSeek V3.1	94	80	70	52	43	67.6%
Gemini 2.5 Pro	82	75	68	68	44	67.4%
GPT-4.1 Nano	100	65	63	59	48	66.9%
Stealth: Healer Alpha	100	73	65	62	34	66.8%
o4 Mini High	86	74	73	72	26	66.1%
Arcee AI: Trinity Large (Preview)	100	100	68	47	10	64.8%
Gemma 3 27B	77	77	70	54	42	64.1%
Qwen 3.5 Flash	96	85	42	41	26	58.1%
Z.AI GLM 4.7 Flash	93	64	62	44	16	56.1%
ByteDance Seed 2.0 Mini	87	84	37	24	20	50.5%
GPT-5 Mini	60	56	51	49	35	50.0%
GPT-5	74	69	60	33	5	48.2%
Z.AI GLM 4.7	77	70	58	26	9	48.1%
Gemini 3 Pro (Preview)	100	76	43	17	0	47.3%
Qwen 3.5 9B	100	64	37	21	13	47.0%
ByteDance Seed 2.0 Lite	69	53	52	35	0	41.8%
Qwen 3.5 122B	71	63	34	23	0	38.2%
Mistral NeMO	77	70	42	0	0	37.8%
Qwen 3.5 35B	55	49	39	29	4	35.0%
Inception Mercury	59	54	27	22	7	33.8%
Qwen 3.5 27B	71	35	29	21	0	31.5%
Gemini 3.1 Flash Lite (Preview)	58	52	30	11	0	30.3%
Mistral Small 3.2 24B	6	0	0	0	0	1.3%

genre

▼

Fantasy: entering an ancient ruin

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Gemini 3.1 Pro (Preview)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
ByteDance Seed 1.6	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	100	100.0%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
DeepSeek-V2 Chat	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
Hermes 3 405B	100	100	100	100	100	100.0%
GPT-5 Nano	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
Qwen 3 32B	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Mistral Large	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
GPT-4o Mini (temp=1)	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Qwen 2.5 72B	100	100	100	100	100	100.0%
Cohere Command R+ (Aug. 2024)	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
LFM2 24B	100	100	100	100	100	100.0%
DeepSeek V3.2	100	100	100	100	99	99.7%
GPT-4o, May 13th (temp=1)	100	100	100	100	99	99.7%
GPT-4o Mini (temp=0)	100	100	100	100	98	99.7%
GPT-4.1	100	100	100	100	98	99.6%
Claude Sonnet 4	100	100	100	100	97	99.5%
Rocinante 12B	100	100	100	100	97	99.4%
Ministral 3 14B	100	100	100	100	96	99.2%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	96	99.2%
Writer: Palmyra X5	100	100	100	100	95	99.1%
Gemini 2.5 Flash	100	100	100	100	95	99.0%
Hermes 3 70B	100	100	100	100	94	98.8%
GPT-5.1	100	100	100	100	94	98.7%
Z.AI GLM 5	100	100	100	100	92	98.4%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	96	96	98.4%
DeepSeek V3 (2024-12-26)	100	100	100	100	91	98.1%
GPT-5.4 Nano (Reasoning)	100	100	100	100	89	97.9%
Inception Mercury 2	100	100	100	100	88	97.5%
Claude 3.5 Sonnet	100	100	100	100	87	97.3%
GPT-5.2	100	100	100	96	90	97.1%
GPT-5 Mini	100	100	99	98	87	96.7%
MiniMax M2.5	100	100	100	100	84	96.7%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	82	96.3%
Gemma 3 27B	100	100	100	100	81	96.3%
o4 Mini	100	100	100	94	87	96.2%
Mistral Small 4 (Reasoning)	100	100	100	93	88	96.1%
Z.AI GLM 4.7	100	100	100	99	82	96.1%
ByteDance Seed 1.6 Flash	100	100	100	100	79	95.9%
Gemma 3 12B	100	100	100	100	79	95.8%
Ministral 8B	100	100	100	89	88	95.5%
GPT-4.1 Mini	100	100	100	100	75	95.0%
MoonshotAI: Kimi K2.5	100	100	100	93	83	95.0%
Gemini 3 Flash (Preview)	100	100	92	92	90	94.8%
Ministral 3 3B	100	100	100	92	82	94.7%
GPT-5.4 Nano	100	100	98	91	82	94.1%
Ministral 3 8B	100	100	100	95	73	93.7%
MiniMax M2.7	100	100	97	92	78	93.5%
Gemini 2.5 Flash (Reasoning)	100	100	100	84	82	93.2%
Claude 3 Haiku	100	100	100	100	65	93.0%
Mistral Small Creative	100	100	94	87	84	93.0%
Stealth: Healer Alpha	100	100	100	88	77	92.9%
Llama 3.1 Nemotron 70B	100	100	100	100	63	92.6%
Arcee AI: Trinity Large (Preview)	100	100	100	100	63	92.5%
Gemini 3 Flash (Preview, Reasoning)	100	100	97	89	74	92.1%
GPT-5	100	94	93	90	83	92.1%
GPT-4o, May 13th (temp=0)	100	100	100	96	63	91.9%
Aion 2.0	100	100	87	87	85	91.7%
Gemini 2.5 Flash Lite	100	96	92	87	84	91.6%
Stealth: Aurora Alpha	100	100	100	88	70	91.5%
Ministral 3B	100	100	100	100	57	91.4%
Claude Opus 4.6 (Reasoning)	100	100	100	100	57	91.3%
Nemotron 3 Super	100	100	95	85	75	91.0%
o4 Mini High	99	98	89	84	84	90.8%
WizardLM 2 8x22b	100	100	100	82	72	90.8%
Stealth: Hunter Alpha	100	99	92	87	74	90.3%
Gemini 3 Pro (Preview)	100	95	92	89	67	88.8%
Nemotron 3 Nano	100	100	90	90	58	87.6%
Claude Sonnet 4.6	100	100	88	78	71	87.4%
Arcee AI: Trinity Mini	100	100	82	77	76	87.0%
Gemma 3 4B	100	100	96	70	69	87.0%
DeepSeek V3.1	100	91	87	81	71	85.8%
Qwen 3.5 397B A17B	100	100	88	72	68	85.6%
Gemini 2.5 Pro	96	93	87	81	70	85.3%
Qwen 3.5 27B	100	100	90	67	66	84.6%
Qwen 3.5 Flash	100	93	86	73	62	82.7%
Qwen 3.5 122B	99	85	80	75	71	81.8%
GPT-4.1 Nano	96	87	87	72	67	81.6%
Z.AI GLM 4.6	95	93	81	60	56	77.2%
Claude Opus 4.6	100	84	81	76	36	75.5%
ByteDance Seed 2.0 Mini	99	99	61	56	56	74.3%
Qwen 3.5 35B	100	79	69	62	60	73.9%
GPT-4o, Aug. 6th (temp=0)	100	82	80	73	32	73.4%
ByteDance Seed 2.0 Lite	100	95	70	70	29	72.7%
Claude Sonnet 4.6 (Reasoning)	100	94	67	58	34	70.5%
Gemini 3.1 Flash Lite (Preview)	92	92	79	47	43	70.4%
Gemini 2.5 Flash Lite (Reasoning)	85	83	72	58	52	69.9%
Inception Mercury	100	100	74	68	0	68.3%
Qwen 3.5 9B	100	79	71	60	16	65.2%
Z.AI GLM 4.7 Flash	89	77	55	54	44	63.9%
Mistral NeMO	49	45	36	24	16	34.1%
Mistral Small 3.2 24B	72	39	39	0	0	30.0%

▼

Horror: alone in an eerie place at night

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	99	99.9%
Claude 3.7 Sonnet	100	100	100	100	98	99.6%
Z.AI GLM 5 Turbo	100	100	100	100	98	99.6%
Grok 4.20 (Beta)	100	100	100	100	97	99.5%
Grok 4	100	100	100	100	97	99.4%
Rocinante 12B	100	100	100	100	96	99.0%
GPT-5.4 Mini	100	100	100	98	97	99.0%
GPT-4.1 Mini	100	100	100	100	94	98.8%
GPT-5.4 Mini (Reasoning)	100	100	100	98	94	98.4%
Qwen3 235B A22B Instruct 2507	100	100	99	99	93	98.2%
Writer: Palmyra X5	100	100	100	100	90	98.0%
Claude 3 Haiku	100	100	100	100	80	96.0%
Hermes 3 405B	100	100	96	96	87	95.7%
Arcee AI: Trinity Large (Preview)	100	100	100	93	84	95.3%
LFM2 24B	100	100	100	91	85	95.1%
Claude Haiku 4.5	100	100	100	100	74	94.8%
GPT-5.4 Mini (Reasoning, Low)	100	99	96	95	83	94.6%
GPT-4o Mini (temp=1)	100	100	100	88	85	94.6%
Mistral Small Creative	100	95	94	91	89	93.6%
Ministral 3B	100	100	100	94	74	93.6%
Mistral Medium 3.1	100	100	100	100	64	92.9%
Qwen 3 32B	100	100	100	100	64	92.7%
Hermes 3 70B	100	100	100	87	77	92.7%
MoonshotAI: Kimi K2.5	100	100	97	95	71	92.4%
Ministral 3 8B	100	100	98	85	75	91.6%
Mistral Large	100	100	95	88	71	90.7%
ByteDance Seed 1.6 Flash	100	100	98	83	71	90.4%
Mistral Large 2	100	100	90	89	70	89.9%
Claude Opus 4.6 (Reasoning)	100	100	100	77	72	89.8%
Grok 4.20 (Beta, Reasoning)	100	100	86	81	80	89.5%
WizardLM 2 8x22b	100	100	100	82	64	89.2%
Cohere Command R+ (Aug. 2024)	100	100	100	73	69	88.4%
GPT-4o, Aug. 6th (temp=0)	100	100	100	85	55	87.9%
Ministral 3 14B	94	92	91	80	78	87.1%
Ministral 8B	100	94	90	80	68	86.4%
Qwen 2.5 72B	100	100	100	69	62	86.3%
GPT-5.1	93	90	89	84	70	85.2%
GPT-5.4 Nano (Reasoning, Low)	95	87	86	84	73	85.0%
Arcee AI: Trinity Mini	100	100	84	72	68	84.8%
Z.AI GLM 5	100	88	82	78	76	84.7%
GPT-4o, May 13th (temp=1)	100	100	94	69	60	84.5%
Mistral Large 3	100	100	85	72	65	84.5%
DeepSeek V3 (2024-12-26)	100	100	75	72	66	82.5%
Mistral Small 4 (Reasoning)	100	100	84	68	58	81.9%
GPT-5 Nano	100	91	89	67	61	81.5%
GPT-5.2	88	85	83	80	68	80.8%
Nemotron 3 Nano	93	88	81	75	64	80.3%
DeepSeek V3.2	84	81	80	79	76	80.0%
Inception Mercury 2	100	92	79	70	58	79.8%
Gemini 3.1 Pro (Preview)	100	84	78	78	57	79.4%
Stealth: Aurora Alpha	99	96	86	66	50	79.1%
GPT-4.1	93	88	77	75	62	78.9%
Mistral Small 4	100	94	78	66	56	78.8%
MiniMax M2.5	99	92	82	66	46	77.1%
GPT-4o Mini (temp=0)	100	93	81	60	51	77.0%
Gemma 3 12B	100	85	67	65	58	75.2%
MiniMax M2.7	100	100	64	55	54	74.5%
Gemma 3 27B	92	89	85	75	31	74.5%
Qwen 3.5 Plus (2026-02-15)	88	74	73	69	68	74.4%
Stealth: Healer Alpha	92	88	72	65	52	73.8%
GPT-5.4 Nano	89	78	74	65	62	73.6%
DeepSeek-V2 Chat	100	100	100	55	11	73.3%
GPT-5 Mini	95	78	71	62	52	71.7%
GPT-5.4 Nano (Reasoning)	78	76	74	65	65	71.6%
Claude Sonnet 4.6	80	75	73	72	56	71.2%
Claude Opus 4.6	89	81	75	56	53	70.6%
Gemini 2.5 Flash	83	76	71	69	49	69.5%
Stealth: Hunter Alpha	72	70	70	67	52	66.1%
Ministral 3 3B	92	78	76	58	26	66.1%
o4 Mini High	90	67	60	50	37	60.6%
Nemotron 3 Super	85	68	60	49	40	60.6%
Gemini 2.5 Flash (Reasoning)	89	73	54	47	38	60.1%
Z.AI GLM 4.7	73	69	61	56	39	59.8%
Claude Sonnet 4.6 (Reasoning)	75	65	65	57	31	58.5%
Aion 2.0	73	64	55	54	46	58.4%
Inception Mercury	85	81	69	57	0	58.2%
Qwen 3.5 397B A17B	69	60	55	53	51	57.7%
Z.AI GLM 4.6	67	61	57	48	47	55.8%
DeepSeek V3.1	71	65	59	51	28	54.7%
Gemma 3 4B	72	67	64	36	33	54.4%
Gemini 3 Flash (Preview, Reasoning)	82	72	56	37	22	53.7%
Gemini 3 Pro (Preview)	68	60	56	48	37	53.5%
o4 Mini	80	60	59	49	12	52.0%
GPT-4.1 Nano	64	59	58	43	32	51.1%
Gemini 2.5 Flash Lite	82	53	42	41	38	51.0%
Gemini 2.5 Flash Lite (Reasoning)	63	49	47	40	39	47.5%
Z.AI GLM 4.7 Flash	61	59	43	41	27	46.0%
Gemini 2.5 Pro	62	52	49	41	25	45.8%
Qwen 3.5 Flash	58	53	52	50	9	44.3%
Gemini 3 Flash (Preview)	62	46	42	36	34	43.9%
GPT-5	67	50	37	24	23	40.1%
Qwen 3.5 122B	52	42	36	34	33	39.2%
ByteDance Seed 1.6	73	49	34	27	12	38.9%
GPT-4o, May 13th (temp=0)	56	51	47	18	18	38.0%
Qwen 3.5 35B	58	37	32	30	25	36.4%
Mistral NeMO	57	42	39	25	17	36.0%
ByteDance Seed 2.0 Mini	54	50	49	11	0	32.8%
Qwen 3.5 9B	49	41	38	21	10	31.8%
ByteDance Seed 2.0 Lite	65	40	30	13	0	29.5%
Qwen 3.5 27B	59	42	28	7	2	27.7%
Gemini 3.1 Flash Lite (Preview)	41	40	35	10	5	25.9%
Mistral Small 3.2 24B	35	0	0	0	0	7.0%

▼

Literary fiction: old friends reunite

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
GPT-5.1	100	100	100	100	95	99.1%
Z.AI GLM 5	100	100	99	96	83	95.5%
Gemma 3 12B	100	100	100	100	76	95.2%
Claude 3.5 Haiku	100	100	100	100	69	93.8%
GPT-5.2	100	100	100	97	60	91.5%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	72	71	88.6%
GPT-4o, May 13th (temp=1)	100	100	100	100	43	88.6%
Llama 3.1 8B	100	100	100	80	63	88.6%
Claude Opus 4.6 (Reasoning)	95	93	92	90	72	88.4%
GPT-5.4 (Reasoning)	100	100	100	74	59	86.5%
Grok 4.20 (Beta, Reasoning)	100	93	91	74	74	86.4%
Llama 3.1 70B	100	100	100	100	27	85.4%
Llama 3.1 Nemotron 70B	100	90	82	78	75	84.9%
Mistral Large	100	100	75	73	66	82.9%
Claude Opus 4	100	100	83	83	40	81.3%
DeepSeek V3 (2025-03-24)	100	100	100	100	6	81.3%
GPT-5.4 Nano (Reasoning)	91	84	81	76	63	79.0%
Arcee AI: Trinity Large (Preview)	100	83	71	70	66	78.1%
GPT-5.4 (Reasoning, Low)	100	100	71	60	56	77.5%
Hermes 3 70B	100	100	90	80	13	76.6%
WizardLM 2 8x22b	100	98	87	62	35	76.5%
MiniMax M2.7	100	100	100	82	0	76.4%
Qwen 2.5 72B	100	100	73	59	50	76.3%
Mistral Large 2	100	100	100	56	25	76.1%
Claude Sonnet 4.5	100	100	100	67	11	75.5%
Claude Opus 4.5	100	100	72	62	38	74.3%
Claude Opus 4.6	100	100	93	56	22	74.2%
Mistral Small 3.2 24B	100	100	100	52	14	73.2%
GPT-5.4 Mini	100	82	70	58	54	72.8%
Grok 4.1 Fast	100	92	78	64	29	72.5%
Claude Sonnet 4	100	81	69	57	50	71.4%
GPT-5.4 Nano (Reasoning, Low)	100	71	69	66	42	69.7%
Z.AI GLM 4.5	94	94	68	57	35	69.5%
Claude Haiku 4.5	100	79	78	49	13	63.5%
Ministral 3B	100	65	61	54	37	63.5%
Grok 4.20 (Beta)	74	71	58	57	56	63.2%
Claude 3.7 Sonnet	100	65	59	43	38	61.1%
MoonshotAI: Kimi K2.5	100	72	57	43	30	60.3%
Cohere Command R+ (Aug. 2024)	100	74	55	51	20	60.0%
Claude Sonnet 4.6	100	99	95	0	0	58.9%
GPT-5.4 Nano	70	64	59	51	45	57.9%
Grok 4 Fast	100	74	69	41	4	57.7%
Nemotron 3 Nano	91	58	49	47	43	57.5%
Writer: Palmyra X5	100	100	71	14	0	57.1%
Claude 3 Haiku	100	60	47	45	30	56.5%
Claude 3.5 Sonnet	100	100	41	37	2	56.1%
Hermes 3 405B	84	67	57	35	33	55.1%
Qwen 3.5 Plus (2026-02-15)	90	71	43	39	29	54.4%
Mistral Small 4	90	53	52	47	29	54.2%
Gemini 3 Pro (Preview)	75	72	64	43	13	53.3%
GPT-5	100	58	57	51	0	53.3%
Mistral Medium 3.1	88	85	64	16	11	53.0%
Rocinante 12B	100	82	72	8	0	52.3%
Gemini 3 Flash (Preview, Reasoning)	82	61	50	39	28	52.1%
MiniMax M2.5	100	93	27	14	13	49.5%
GPT-5.4	63	53	52	40	39	49.5%
Claude Sonnet 4.6 (Reasoning)	73	54	47	46	23	48.6%
Mistral NeMO	100	100	22	20	0	48.4%
DeepSeek V3 (2024-12-26)	71	60	57	29	22	48.0%
GPT-5 Mini	73	73	67	26	0	47.9%
Z.AI GLM 5 Turbo	92	78	56	14	0	47.8%
GPT-4o, Aug. 6th (temp=1)	100	51	33	30	23	47.5%
Grok 4	70	55	49	33	20	45.2%
LFM2 24B	68	52	43	32	27	44.7%
Gemini 2.5 Flash (Reasoning)	100	81	26	16	0	44.5%
GPT-4o Mini (temp=1)	100	47	39	29	7	44.3%
GPT-4o, May 13th (temp=0)	100	77	42	0	0	43.7%
Mistral Small Creative	66	59	45	37	7	42.8%
GPT-4o, Aug. 6th (temp=0)	90	78	40	0	0	41.6%
Mistral Large 3	72	58	52	23	0	41.0%
Ministral 3 3B	100	45	30	16	13	41.0%
ByteDance Seed 2.0 Lite	63	49	46	38	8	40.9%
ByteDance Seed 1.6 Flash	100	65	28	10	0	40.5%
Gemini 3.1 Pro (Preview)	66	52	41	41	0	40.0%
Stealth: Healer Alpha	91	75	32	0	0	39.5%
ByteDance Seed 2.0 Mini	89	74	23	10	0	39.0%
Ministral 3 14B	54	51	49	24	17	38.9%
Mistral Small 4 (Reasoning)	58	57	43	32	0	37.9%
Ministral 8B	51	49	45	36	0	36.1%
ByteDance Seed 1.6	77	53	33	5	0	33.5%
Gemini 2.5 Flash Lite	80	53	11	10	5	31.7%
Ministral 3 8B	55	53	28	14	5	31.2%
GPT-4o Mini (temp=0)	100	27	20	8	0	31.0%
Qwen3 235B A22B Instruct 2507	64	59	20	11	0	30.9%
Gemini 2.5 Flash Lite (Reasoning)	60	52	34	3	0	29.7%
Qwen 3 32B	76	20	18	16	16	29.2%
DeepSeek-V2 Chat	100	26	6	0	0	26.4%
Qwen 3.5 35B	67	45	16	0	0	25.7%
Gemini 3 Flash (Preview)	39	32	29	27	0	25.6%
GPT-4.1	82	33	10	0	0	25.1%
Nemotron 3 Super	48	34	33	9	0	24.9%
GPT-5 Nano	30	30	29	21	11	24.0%
GPT-4.1 Mini	69	20	12	9	7	23.4%
Qwen 3.5 27B	71	37	6	0	0	22.6%
Arcee AI: Trinity Mini	60	33	20	0	0	22.5%
Stealth: Hunter Alpha	61	31	11	3	0	21.1%
Qwen 3.5 397B A17B	55	28	18	0	0	20.4%
Aion 2.0	46	20	4	0	0	13.9%
Inception Mercury 2	33	29	6	0	0	13.7%
Stealth: Aurora Alpha	38	14	12	3	0	13.4%
Gemma 3 27B	30	22	11	0	0	12.7%
Gemini 2.5 Flash	32	20	3	0	0	10.9%
Z.AI GLM 4.6	20	18	14	0	0	10.4%
Z.AI GLM 4.7 Flash	51	0	0	0	0	10.3%
Qwen 3.5 Flash	51	0	0	0	0	10.1%
Qwen 3.5 9B	40	4	0	0	0	8.6%
Inception Mercury	19	17	0	0	0	7.2%
Gemini 2.5 Pro	23	5	0	0	0	5.6%
DeepSeek V3.2	23	0	0	0	0	4.6%
GPT-4.1 Nano	16	6	0	0	0	4.4%
Gemini 3.1 Flash Lite (Preview)	20	0	0	0	0	4.0%
Z.AI GLM 4.7	16	0	0	0	0	3.2%
o4 Mini	10	4	0	0	0	2.7%
Qwen 3.5 122B	3	0	0	0	0	0.6%
o4 Mini High	0	0	0	0	0	0.0%
DeepSeek V3.1	0	0	0	0	0	0.0%
Gemma 3 4B	0	0	0	0	0	0.0%

▼

Mystery: examining a crime scene

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
MoonshotAI: Kimi K2.5	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	100	100.0%
GPT-4o, May 13th (temp=0)	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=0)	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Mistral Large	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Ministral 3 14B	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Ministral 8B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Ministral 3B	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	99.9%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	99.9%
Qwen 2.5 72B	100	100	100	100	99	99.8%
Cohere Command R+ (Aug. 2024)	100	100	100	100	99	99.8%
Llama 3.1 70B	100	100	100	100	99	99.8%
GPT-5.2	100	100	100	99	96	98.9%
Gemma 3 12B	100	100	100	100	94	98.9%
GPT-5.4 Nano	100	100	100	100	94	98.8%
Z.AI GLM 5	100	100	100	98	94	98.4%
Nemotron 3 Nano	100	100	100	100	91	98.3%
Z.AI GLM 4.5	100	100	100	100	91	98.2%
Claude 3.5 Haiku	100	100	100	100	91	98.2%
GPT-4o, May 13th (temp=1)	100	100	100	100	91	98.2%
Mistral Small 4 (Reasoning)	100	100	100	100	90	98.0%
GPT-5.4 Nano (Reasoning)	100	100	100	96	94	98.0%
Gemini 3.1 Pro (Preview)	100	100	100	100	88	97.6%
ByteDance Seed 1.6 Flash	100	100	100	100	88	97.6%
Aion 2.0	100	100	100	94	94	97.5%
Hermes 3 405B	100	100	100	100	87	97.3%
Qwen3 235B A22B Instruct 2507	100	100	100	100	85	97.0%
Writer: Palmyra X5	100	100	100	94	91	97.0%
Inception Mercury 2	100	100	100	100	85	97.0%
GPT-4.1 Mini	100	100	100	100	84	96.9%
Mistral Small Creative	100	100	100	100	82	96.4%
MiniMax M2.5	100	100	100	98	83	96.3%
Gemma 3 4B	100	100	100	91	91	96.3%
DeepSeek-V2 Chat	100	100	100	93	87	96.0%
WizardLM 2 8x22b	100	100	97	95	87	95.8%
Arcee AI: Trinity Large (Preview)	100	100	100	91	87	95.6%
GPT-5 Nano	100	100	100	91	85	95.3%
DeepSeek V3.2	100	100	97	92	87	95.2%
DeepSeek V3 (2024-12-26)	100	100	100	100	76	95.2%
Claude Opus 4.6 (Reasoning)	100	100	98	90	88	95.1%
Gemini 2.5 Flash (Reasoning)	100	100	100	100	75	95.0%
ByteDance Seed 2.0 Lite	100	100	95	94	87	95.0%
o4 Mini High	100	100	99	92	83	94.8%
Z.AI GLM 5 Turbo	100	100	100	93	80	94.7%
Ministral 3 8B	100	100	100	88	85	94.6%
GPT-4o Mini (temp=0)	100	100	99	94	78	94.3%
Qwen 3.5 Plus (2026-02-15)	100	100	98	92	81	94.1%
Gemma 3 27B	100	100	100	100	67	93.5%
Claude Sonnet 4.6	100	100	99	98	69	93.4%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	66	93.2%
Gemini 2.5 Flash	100	100	94	91	78	92.5%
LFM2 24B	100	100	100	100	62	92.3%
Qwen 3 32B	100	100	100	100	62	92.3%
GPT-5 Mini	100	100	89	88	81	91.6%
Gemini 2.5 Flash Lite	100	100	100	89	68	91.5%
Grok 4.20 (Beta)	100	100	95	89	70	90.9%
Mistral Large 3	100	100	89	88	71	89.5%
Nemotron 3 Super	100	91	87	85	85	89.4%
Gemini 2.5 Flash Lite (Reasoning)	100	99	85	83	78	88.9%
GPT-4.1	100	100	100	81	63	88.8%
Ministral 3 3B	100	100	100	100	44	88.8%
Qwen 3.5 397B A17B	97	92	90	86	76	88.1%
Gemini 2.5 Pro	100	96	94	79	69	87.7%
MiniMax M2.7	100	100	93	82	62	87.4%
Gemini 3 Pro (Preview)	100	99	87	79	72	87.2%
Mistral Medium 3.1	100	91	82	81	80	86.7%
GPT-5.1	100	87	83	80	78	85.7%
Claude Opus 4.6	100	98	92	76	59	84.8%
Stealth: Healer Alpha	96	88	87	77	75	84.5%
Rocinante 12B	100	100	100	65	56	84.3%
GPT-4o Mini (temp=1)	100	100	89	76	53	83.6%
o4 Mini	91	91	79	77	76	82.9%
ByteDance Seed 1.6	100	100	82	75	58	82.9%
Stealth: Hunter Alpha	100	100	99	57	50	81.1%
DeepSeek V3.1	100	98	91	73	42	80.8%
Hermes 3 70B	100	100	100	68	28	79.2%
ByteDance Seed 2.0 Mini	100	100	87	54	53	78.7%
Z.AI GLM 4.6	100	93	75	61	59	77.6%
Z.AI GLM 4.7	90	90	79	70	58	77.5%
Gemini 3 Flash (Preview)	100	98	76	71	36	76.1%
Z.AI GLM 4.7 Flash	99	87	79	59	46	73.9%
Claude Sonnet 4.6 (Reasoning)	100	81	77	63	41	72.4%
Stealth: Aurora Alpha	100	93	86	62	0	68.1%
GPT-5	80	79	72	70	38	67.7%
Gemini 3 Flash (Preview, Reasoning)	100	77	70	48	30	64.8%
Qwen 3.5 Flash	89	64	62	56	31	60.3%
Qwen 3.5 9B	100	62	54	42	41	59.9%
GPT-4.1 Nano	78	70	61	50	28	57.4%
Qwen 3.5 122B	75	74	45	41	41	55.2%
Mistral Small 3.2 24B	100	92	53	31	0	55.1%
Gemini 3.1 Flash Lite (Preview)	79	54	47	42	32	51.0%
Qwen 3.5 35B	64	61	51	50	29	50.8%
Arcee AI: Trinity Mini	62	52	49	44	36	48.5%
Mistral NeMO	100	44	38	31	7	44.0%
Qwen 3.5 27B	55	52	38	33	20	39.7%
Inception Mercury	100	69	15	0	0	36.7%

▼

Romance: separated couple reunites

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Llama 3.1 8B	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	98	99.6%
Mistral Large	100	100	100	100	89	97.7%
GPT-5.4 Mini	100	100	96	94	91	96.1%
Llama 3.1 70B	100	100	98	91	90	95.9%
Claude 3.5 Haiku	100	100	93	79	66	87.5%
Mistral Large 3	100	98	93	74	54	83.8%
GPT-5.4 Mini (Reasoning)	100	100	80	72	62	82.8%
GPT-5.4 Mini (Reasoning, Low)	100	86	79	79	69	82.5%
DeepSeek V3 (2025-03-24)	100	100	84	72	53	81.9%
GPT-5.4 Nano (Reasoning, Low)	100	94	89	65	58	81.1%
Rocinante 12B	100	100	78	65	61	81.0%
Ministral 3 3B	100	83	75	75	68	80.0%
Mistral Small Creative	100	100	100	60	35	79.1%
Mistral Small 4 (Reasoning)	97	95	67	60	55	74.8%
GPT-5.4	92	84	74	62	52	72.8%
Claude 3 Haiku	99	78	62	58	54	70.3%
GPT-5.4 Nano (Reasoning)	95	91	74	61	29	70.1%
Ministral 3B	95	82	69	52	51	69.5%
Mistral Large 2	100	100	76	41	23	68.2%
Qwen 2.5 72B	99	61	61	60	54	67.2%
GPT-5.4 Nano	80	72	60	56	52	63.8%
GPT-5.4 (Reasoning)	87	68	62	56	45	63.5%
Grok 4 Fast	75	72	62	55	53	63.4%
Ministral 3 8B	100	89	45	39	38	62.5%
GPT-5.4 (Reasoning, Low)	84	68	64	57	36	61.9%
GPT-4o, May 13th (temp=1)	100	87	75	39	0	60.0%
GPT-5.2	77	57	55	54	51	58.9%
Grok 4.1 Fast	70	66	56	49	29	54.1%
Grok 4.20 (Beta, Reasoning)	71	59	49	49	39	53.4%
Z.AI GLM 5 Turbo	80	74	54	49	0	51.4%
Qwen 3 32B	100	60	45	38	14	51.4%
Ministral 8B	84	67	58	25	22	51.1%
Mistral Small 4	100	72	41	26	13	50.6%
MiniMax M2.7	59	57	47	47	42	50.5%
LFM2 24B	95	63	50	24	17	50.0%
Claude 3.7 Sonnet	68	58	57	42	24	49.9%
Grok 4.20 (Beta)	72	54	49	46	21	48.5%
Gemma 3 27B	100	100	42	0	0	48.3%
Claude Sonnet 4.5	100	47	35	30	25	47.4%
WizardLM 2 8x22b	78	72	44	20	20	46.7%
Z.AI GLM 5	67	63	59	32	9	46.1%
Mistral Small 3.2 24B	65	58	58	35	12	45.7%
Claude Opus 4.5	58	58	57	47	7	45.5%
Claude Haiku 4.5	78	44	42	37	23	44.9%
Mistral Medium 3.1	75	73	38	35	2	44.6%
Claude Sonnet 4	94	54	34	31	10	44.6%
Hermes 3 405B	100	67	42	12	0	44.1%
Ministral 3 14B	64	57	41	33	20	42.8%
Qwen 3.5 Plus (2026-02-15)	83	61	34	26	9	42.3%
Cohere Command R+ (Aug. 2024)	75	73	31	12	4	38.7%
Hermes 3 70B	100	93	0	0	0	38.6%
MoonshotAI: Kimi K2.5	56	55	54	22	0	37.3%
Claude Opus 4.6	76	63	26	6	0	34.2%
Mistral NeMO	77	49	41	3	0	34.0%
Gemini 3 Flash (Preview)	54	35	31	22	22	32.8%
Claude Opus 4	67	66	22	1	0	31.3%
ByteDance Seed 1.6	81	42	33	0	0	31.0%
Gemini 3 Flash (Preview, Reasoning)	68	60	23	1	0	30.4%
Z.AI GLM 4.5	51	47	36	9	0	28.5%
Gemini 3.1 Pro (Preview)	43	42	27	18	10	28.1%
DeepSeek V3 (2024-12-26)	89	20	14	8	0	26.1%
GPT-5 Nano	67	30	25	3	0	24.9%
GPT-4o, Aug. 6th (temp=1)	87	35	0	0	0	24.3%
DeepSeek V3.2	65	27	18	8	3	24.2%
GPT-4o, Aug. 6th (temp=0)	69	42	0	0	0	22.1%
Arcee AI: Trinity Large (Preview)	90	20	0	0	0	22.1%
Grok 4	56	16	14	13	10	22.0%
GPT-4o, May 13th (temp=0)	85	20	0	0	0	21.0%
Qwen3 235B A22B Instruct 2507	46	39	14	1	0	20.1%
MiniMax M2.5	50	28	14	4	0	19.2%
GPT-4.1 Mini	31	29	17	11	0	17.6%
DeepSeek-V2 Chat	57	15	14	0	0	17.3%
Writer: Palmyra X5	33	30	23	0	0	17.1%
Gemini 2.5 Flash	51	18	16	0	0	17.1%
Claude Opus 4.6 (Reasoning)	32	28	22	2	0	16.8%
Claude 3.5 Sonnet	27	24	15	11	0	15.5%
GPT-4.1	23	20	20	14	0	15.3%
GPT-5 Mini	31	27	10	8	0	15.2%
o4 Mini	35	30	10	0	0	15.0%
Gemini 2.5 Pro	22	20	18	8	4	14.5%
Gemma 3 12B	46	23	2	0	0	14.2%
Stealth: Aurora Alpha	31	27	13	0	0	14.2%
Qwen 3.5 397B A17B	30	24	17	0	0	14.2%
Stealth: Healer Alpha	33	28	9	0	0	14.0%
GPT-5.1	29	19	11	11	0	13.9%
Gemini 2.5 Flash Lite	32	20	12	0	0	12.9%
GPT-4o Mini (temp=1)	20	20	10	7	6	12.5%
Inception Mercury	35	27	0	0	0	12.5%
ByteDance Seed 1.6 Flash	60	0	0	0	0	12.0%
Nemotron 3 Nano	26	25	6	0	0	11.2%
Stealth: Hunter Alpha	24	13	12	7	0	11.2%
Gemini 3 Pro (Preview)	35	12	8	0	0	10.9%
Z.AI GLM 4.7	28	24	0	0	0	10.3%
Aion 2.0	42	8	1	0	0	10.3%
Qwen 3.5 122B	22	12	9	3	0	9.2%
GPT-4.1 Nano	35	9	0	0	0	8.8%
Claude Sonnet 4.6	38	4	2	0	0	8.6%
ByteDance Seed 2.0 Lite	41	0	0	0	0	8.1%
Gemini 2.5 Flash Lite (Reasoning)	20	11	9	0	0	8.0%
Arcee AI: Trinity Mini	24	16	0	0	0	8.0%
DeepSeek V3.1	18	18	4	0	0	8.0%
Qwen 3.5 Flash	35	0	0	0	0	7.1%
Qwen 3.5 27B	22	7	0	0	0	5.7%
GPT-4o Mini (temp=0)	28	0	0	0	0	5.6%
Inception Mercury 2	20	1	0	0	0	4.2%
GPT-5	20	0	0	0	0	4.0%
Qwen 3.5 9B	11	0	0	0	0	2.2%
Z.AI GLM 4.6	9	1	0	0	0	2.1%
Gemma 3 4B	5	0	0	0	0	1.0%
Qwen 3.5 35B	3	0	0	0	0	0.6%
Nemotron 3 Super	1	1	0	0	0	0.5%
Claude Sonnet 4.6 (Reasoning)	0	0	0	0	0	0.0%
o4 Mini High	0	0	0	0	0	0.0%
ByteDance Seed 2.0 Mini	0	0	0	0	0	0.0%
Gemini 2.5 Flash (Reasoning)	0	0	0	0	0	0.0%
Gemini 3.1 Flash Lite (Preview)	0	0	0	0	0	0.0%
Z.AI GLM 4.7 Flash	0	0	0	0	0	0.0%

▼

Thriller: chase through city streets

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
DeepSeek V3 (2024-12-26)	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4.1 Mini	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	99	99.8%
Claude Sonnet 4	100	100	100	100	99	99.7%
Claude Opus 4	100	100	100	100	98	99.7%
GPT-5.4 (Reasoning, Low)	100	100	100	100	98	99.7%
GPT-5.4	100	100	100	100	97	99.3%
ByteDance Seed 1.6 Flash	100	100	100	98	97	98.9%
Mistral Small Creative	100	100	98	98	95	98.2%
Mistral Large 3	100	100	100	100	91	98.2%
Mistral Large	100	100	100	100	90	98.1%
Claude Haiku 4.5	100	100	100	96	93	97.8%
Claude Opus 4.5	100	100	100	100	87	97.3%
Grok 4.20 (Beta, Reasoning)	100	100	100	94	92	97.0%
Mistral Medium 3.1	100	100	97	94	93	96.8%
Grok 4.1 Fast	100	100	99	95	89	96.5%
Gemini 2.5 Flash	100	100	100	100	79	95.8%
Rocinante 12B	100	100	100	100	79	95.8%
Grok 4	100	100	100	96	82	95.7%
GPT-5.2	100	100	98	91	89	95.7%
Arcee AI: Trinity Large (Preview)	100	100	100	100	79	95.7%
Z.AI GLM 5 Turbo	100	100	100	89	88	95.5%
DeepSeek V3 (2025-03-24)	100	100	100	100	76	95.3%
Mistral Large 2	100	100	100	100	75	95.0%
Mistral Small 4 (Reasoning)	100	100	100	99	74	94.6%
Ministral 3 14B	100	100	100	98	73	94.3%
Cohere Command R+ (Aug. 2024)	100	100	95	92	80	93.4%
Qwen 3 32B	100	100	99	87	73	91.9%
Writer: Palmyra X5	100	100	98	92	67	91.3%
GPT-5.4 Mini (Reasoning, Low)	100	100	97	84	73	90.9%
GPT-5 Nano	100	100	100	81	73	90.9%
LFM2 24B	100	100	100	89	64	90.6%
MiniMax M2.5	100	100	88	83	82	90.6%
Ministral 3 8B	100	100	99	98	55	90.3%
MiniMax M2.7	100	91	85	85	85	89.4%
GPT-5.1	100	99	94	86	67	89.0%
Z.AI GLM 5	100	100	94	80	69	88.5%
Arcee AI: Trinity Mini	100	100	100	76	65	88.3%
GPT-5.4 Nano (Reasoning, Low)	100	90	85	82	82	87.8%
DeepSeek-V2 Chat	100	100	91	84	62	87.3%
GPT-4o, Aug. 6th (temp=0)	97	97	91	87	63	87.0%
Hermes 3 405B	100	100	100	75	55	86.0%
Qwen3 235B A22B Instruct 2507	100	100	94	83	52	85.8%
Nemotron 3 Super	100	91	81	80	77	85.8%
Grok 4 Fast	100	100	100	100	26	85.1%
GPT-4o, May 13th (temp=0)	100	100	81	79	64	84.8%
Qwen 2.5 72B	100	98	90	74	60	84.4%
Nemotron 3 Nano	100	90	90	72	67	83.7%
Ministral 8B	100	99	92	67	59	83.3%
MoonshotAI: Kimi K2.5	100	87	87	75	65	82.7%
Gemini 2.5 Flash (Reasoning)	100	92	81	71	68	82.5%
GPT-4o, May 13th (temp=1)	100	91	82	71	67	82.3%
GPT-4o Mini (temp=1)	100	87	81	73	62	80.7%
Gemini 2.5 Flash Lite (Reasoning)	100	96	76	67	63	80.6%
Gemini 3.1 Pro (Preview)	98	93	91	61	58	80.4%
Ministral 3 3B	100	94	79	76	49	79.7%
GPT-5.4 Nano (Reasoning)	85	82	82	77	72	79.6%
Gemma 3 12B	100	97	85	64	49	79.0%
Stealth: Hunter Alpha	100	87	80	67	59	78.5%
GPT-5.4 Nano	97	81	79	75	60	78.3%
Qwen 3.5 Plus (2026-02-15)	100	76	75	72	69	78.2%
Ministral 3B	100	100	75	72	40	77.3%
Gemma 3 4B	100	89	79	66	48	76.4%
GPT-4.1	100	100	63	61	56	76.0%
Aion 2.0	100	92	68	63	52	75.0%
o4 Mini	100	86	81	61	47	75.0%
Mistral Small 4	100	72	68	68	62	74.1%
DeepSeek V3.2	89	79	76	65	59	73.6%
ByteDance Seed 2.0 Lite	100	87	80	73	20	72.0%
Qwen 3.5 397B A17B	84	75	74	63	59	70.9%
GPT-4o, Aug. 6th (temp=1)	89	80	70	70	43	70.3%
GPT-4o Mini (temp=0)	78	73	69	68	61	70.0%
Claude Opus 4.6 (Reasoning)	79	79	72	71	45	69.3%
Stealth: Healer Alpha	95	75	72	71	32	68.9%
Gemini 2.5 Flash Lite	100	79	67	52	47	68.9%
Inception Mercury 2	98	77	58	53	50	67.2%
Z.AI GLM 4.6	100	87	64	57	26	66.8%
Claude Opus 4.6	92	91	61	44	43	66.5%
WizardLM 2 8x22b	100	100	79	53	0	66.3%
o4 Mini High	92	68	67	66	36	66.0%
Gemini 2.5 Pro	85	80	57	52	49	64.5%
GPT-5 Mini	82	70	67	56	41	63.1%
Gemini 3 Flash (Preview, Reasoning)	98	79	50	43	38	61.7%
Gemma 3 27B	81	80	50	42	33	57.3%
DeepSeek V3.1	75	74	59	33	31	54.5%
GPT-4.1 Nano	85	76	60	44	7	54.2%
Gemini 3 Pro (Preview)	61	54	52	51	44	52.5%
ByteDance Seed 2.0 Mini	62	57	52	42	33	49.5%
Qwen 3.5 9B	86	68	56	36	0	49.2%
Gemini 3 Flash (Preview)	62	59	43	40	37	48.1%
Z.AI GLM 4.7	61	57	49	46	20	46.6%
Qwen 3.5 27B	71	65	38	32	24	46.0%
Hermes 3 70B	84	62	57	22	0	45.0%
Stealth: Aurora Alpha	79	77	31	26	0	42.6%
ByteDance Seed 1.6	68	50	47	20	15	39.8%
Z.AI GLM 4.7 Flash	65	51	41	37	5	39.7%
Claude Sonnet 4.6	60	39	35	33	28	38.9%
Claude Sonnet 4.6 (Reasoning)	58	51	42	23	18	38.2%
Gemini 3.1 Flash Lite (Preview)	48	43	39	29	29	37.5%
Qwen 3.5 35B	77	58	41	6	2	36.7%
GPT-5	45	37	30	28	17	31.4%
Qwen 3.5 Flash	45	34	33	23	15	29.8%
Mistral NeMO	55	31	25	20	18	29.8%
Qwen 3.5 122B	62	31	12	9	0	22.6%
Mistral Small 3.2 24B	30	25	0	0	0	11.1%
Inception Mercury	38	8	0	0	0	9.3%

Novelcrafter Default Prompt

▼

Fantasy: entering an ancient ruin

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.6 (Reasoning)	100	100	100	100	100	100.0%
Gemini 3.1 Pro (Preview)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
Claude Sonnet 4.6 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
Qwen 3.5 397B A17B	100	100	100	100	100	100.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
MoonshotAI: Kimi K2.5	100	100	100	100	100	100.0%
ByteDance Seed 1.6	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
Gemini 3 Flash (Preview, Reasoning)	100	100	100	100	100	100.0%
GPT-5.2	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
MiniMax M2.7	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Gemini 2.5 Flash (Reasoning)	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	100	100.0%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	100.0%
Gemini 3 Flash (Preview)	100	100	100	100	100	100.0%
DeepSeek-V2 Chat	100	100	100	100	100	100.0%
ByteDance Seed 2.0 Lite	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
GPT-4o, May 13th (temp=1)	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
DeepSeek V3 (2024-12-26)	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-4.1 Mini	100	100	100	100	100	100.0%
Hermes 3 405B	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=1)	100	100	100	100	100	100.0%
GPT-5 Nano	100	100	100	100	100	100.0%
GPT-4o, Aug. 6th (temp=0)	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
Mistral Small 4 (Reasoning)	100	100	100	100	100	100.0%
Qwen 3 32B	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning)	100	100	100	100	100	100.0%
Mistral Large	100	100	100	100	100	100.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	100	100.0%
Writer: Palmyra X5	100	100	100	100	100	100.0%
Inception Mercury	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	100	100	100.0%
GPT-4o Mini (temp=1)	100	100	100	100	100	100.0%
Gemma 3 12B	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Gemma 3 27B	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Nemotron 3 Nano	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Qwen 2.5 72B	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Mistral Small Creative	100	100	100	100	100	100.0%
Hermes 3 70B	100	100	100	100	100	100.0%
Ministral 3 14B	100	100	100	100	100	100.0%
GPT-4.1 Nano	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Cohere Command R+ (Aug. 2024)	100	100	100	100	100	100.0%
Ministral 8B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Ministral 3B	100	100	100	100	100	100.0%
LFM2 24B	100	100	100	100	100	100.0%
Rocinante 12B	100	100	100	100	100	100.0%
GPT-5.1	100	100	100	100	100	99.9%
Z.AI GLM 5	100	100	100	100	98	99.7%
Ministral 3 3B	100	100	100	100	98	99.5%
Claude Haiku 4.5	100	100	100	100	97	99.5%
Arcee AI: Trinity Large (Preview)	100	100	100	100	96	99.2%
GPT-4o Mini (temp=0)	100	100	100	100	95	99.0%
MiniMax M2.5	100	100	100	100	92	98.5%
Claude Opus 4.6	100	100	100	100	91	98.3%
Claude Sonnet 4	100	100	100	100	91	98.2%
Ministral 3 8B	100	100	100	100	91	98.2%
GPT-4.1	100	100	100	100	91	98.1%
Gemini 2.5 Flash	100	100	100	95	94	97.9%
Qwen 3.5 27B	100	100	100	100	88	97.6%
Gemma 3 4B	100	100	100	100	87	97.3%
GPT-5.4 Nano	100	100	100	99	86	97.1%
Inception Mercury 2	100	100	97	96	91	96.9%
Gemini 2.5 Flash Lite (Reasoning)	100	100	100	100	84	96.7%
DeepSeek V3.2	100	100	100	100	82	96.4%
Stealth: Hunter Alpha	100	100	100	100	80	96.1%
Arcee AI: Trinity Mini	100	99	99	97	83	95.8%
GPT-4o, May 13th (temp=0)	100	100	100	100	77	95.5%
GPT-5 Mini	100	100	100	99	76	95.0%
Gemini 3 Pro (Preview)	100	100	100	96	78	94.8%
Claude Sonnet 4.6	100	100	100	89	84	94.6%
Z.AI GLM 4.7	100	100	100	100	73	94.6%
Stealth: Aurora Alpha	100	100	100	87	85	94.4%
DeepSeek V3.1	100	100	100	87	83	93.9%
o4 Mini	100	100	100	99	65	92.7%
Z.AI GLM 4.7 Flash	100	100	92	91	80	92.7%
Qwen 3.5 122B	100	100	100	96	66	92.5%
GPT-5	100	100	100	93	67	92.1%
Qwen 3.5 Flash	100	100	94	90	75	91.8%
ByteDance Seed 1.6 Flash	100	100	95	88	76	91.7%
WizardLM 2 8x22b	100	100	100	92	65	91.5%
Qwen 3.5 35B	100	100	100	82	75	91.3%
Nemotron 3 Super	100	100	100	85	71	91.3%
Aion 2.0	100	100	94	87	71	90.3%
Stealth: Healer Alpha	100	100	99	79	68	89.3%
Gemini 3.1 Flash Lite (Preview)	100	100	97	97	52	89.2%
o4 Mini High	100	100	93	81	69	88.7%
Gemini 2.5 Flash Lite	100	100	95	73	71	88.0%
Gemini 2.5 Pro	100	98	91	83	61	86.6%
Z.AI GLM 4.6	100	96	75	74	62	81.5%
ByteDance Seed 2.0 Mini	100	76	74	74	73	79.6%
Qwen 3.5 9B	100	82	82	63	45	74.5%
Mistral Small 3.2 24B	100	61	49	43	18	54.2%
Mistral NeMO	100	63	41	32	0	47.3%

▼

Horror: alone in an eerie place at night

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Gemini 3.1 Pro (Preview)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Rocinante 12B	100	100	100	100	100	100.0%
Mistral Large 3	100	100	100	100	100	99.9%
Claude Opus 4.5	100	100	100	100	99	99.8%
Llama 3.1 70B	100	100	100	100	98	99.6%
Claude 3 Haiku	100	100	100	100	97	99.5%
Hermes 3 70B	100	100	100	100	97	99.4%
Mistral Large 2	100	100	100	100	97	99.3%
LFM2 24B	100	100	100	98	98	99.1%
MiniMax M2.7	100	100	100	100	93	98.6%
Z.AI GLM 5	100	100	100	100	91	98.2%
Claude Opus 4.6	100	100	100	98	93	98.1%
Z.AI GLM 4.5	100	100	100	100	90	98.0%
Grok 4.20 (Beta)	100	100	100	100	90	97.9%
Writer: Palmyra X5	100	100	100	95	94	97.8%
Mistral Small 4 (Reasoning)	100	100	100	100	89	97.8%
DeepSeek V3 (2025-03-24)	100	100	100	100	89	97.8%
GPT-5.1	100	100	100	100	87	97.3%
Ministral 3B	100	100	100	100	87	97.3%
Ministral 3 8B	100	100	96	95	92	96.8%
Mistral Large	100	100	100	100	83	96.7%
GPT-5 Nano	100	100	100	91	90	96.3%
Qwen3 235B A22B Instruct 2507	100	100	100	92	88	95.9%
GPT-4o, Aug. 6th (temp=1)	100	100	100	96	83	95.7%
Llama 3.1 Nemotron 70B	100	100	100	100	77	95.3%
GPT-5.4 Mini	100	100	100	100	76	95.1%
Qwen 3 32B	100	100	99	98	77	94.8%
Arcee AI: Trinity Large (Preview)	100	100	100	100	74	94.8%
MiniMax M2.5	100	100	96	94	83	94.6%
Mistral Small Creative	100	100	100	95	75	94.0%
Inception Mercury 2	100	100	100	92	76	93.6%
Claude Opus 4.6 (Reasoning)	100	100	100	93	72	93.1%
ByteDance Seed 1.6 Flash	100	100	100	88	76	92.7%
GPT-5.2	100	100	91	88	84	92.6%
Gemma 3 27B	100	98	96	93	72	91.8%
GPT-5.4 Mini (Reasoning, Low)	100	100	94	92	73	91.6%
GPT-4.1 Mini	100	98	92	85	82	91.4%
Gemma 3 12B	100	100	99	88	68	91.0%
Qwen 3.5 397B A17B	100	100	95	83	74	90.4%
GPT-4o Mini (temp=1)	100	98	94	85	67	88.6%
Ministral 3 14B	100	100	100	85	57	88.5%
Hermes 3 405B	100	100	100	99	42	88.3%
Ministral 8B	100	100	100	77	62	87.8%
GPT-4o, Aug. 6th (temp=0)	100	99	97	92	50	87.5%
GPT-4o, May 13th (temp=1)	100	100	98	76	63	87.3%
Grok 4	100	96	95	77	68	87.1%
Nemotron 3 Nano	97	94	92	79	72	86.8%
Qwen 3.5 Plus (2026-02-15)	100	92	85	84	70	86.3%
GPT-4.1	100	100	98	94	39	86.2%
Cohere Command R+ (Aug. 2024)	100	100	100	77	53	86.1%
GPT-5.4 Mini (Reasoning)	100	100	90	72	68	86.0%
Ministral 3 3B	100	87	83	82	76	85.6%
GPT-5.4 Nano (Reasoning)	100	95	91	70	64	84.2%
DeepSeek V3 (2024-12-26)	100	100	100	64	53	83.3%
Claude Sonnet 4.6 (Reasoning)	98	96	92	67	55	81.8%
GPT-5.4 Nano	90	90	88	83	50	80.2%
Gemini 2.5 Flash	100	83	77	73	68	80.2%
MoonshotAI: Kimi K2.5	100	88	80	67	64	79.8%
Qwen 2.5 72B	100	99	83	70	44	79.3%
Gemini 3 Flash (Preview, Reasoning)	92	89	77	70	58	77.2%
Grok 4.20 (Beta, Reasoning)	92	90	84	75	39	75.9%
Stealth: Aurora Alpha	96	94	88	64	33	75.0%
Stealth: Hunter Alpha	90	77	75	70	63	74.7%
Aion 2.0	92	85	76	61	56	74.0%
Qwen 3.5 Flash	89	88	73	66	50	73.3%
DeepSeek V3.1	100	73	70	70	50	72.6%
GPT-5.4 Nano (Reasoning, Low)	90	72	69	67	65	72.6%
Qwen 3.5 27B	100	77	76	68	31	70.3%
Z.AI GLM 4.6	87	76	72	66	47	69.6%
Gemini 2.5 Flash (Reasoning)	91	86	58	56	51	68.2%
Stealth: Healer Alpha	100	92	64	49	36	68.2%
Gemini 3 Pro (Preview)	76	72	68	67	52	67.0%
Gemini 3 Flash (Preview)	83	82	59	56	51	66.5%
DeepSeek V3.2	93	72	58	53	49	64.8%
Gemini 2.5 Flash Lite	93	73	71	46	41	64.7%
WizardLM 2 8x22b	75	67	66	65	49	64.6%
GPT-4o Mini (temp=0)	90	71	59	57	39	63.3%
Z.AI GLM 4.7	77	67	63	62	44	62.5%
o4 Mini	77	75	66	64	31	62.4%
GPT-5 Mini	95	63	60	53	31	60.2%
GPT-4.1 Nano	75	59	59	54	46	58.5%
Gemini 2.5 Flash Lite (Reasoning)	91	67	49	48	35	58.0%
GPT-4o, May 13th (temp=0)	77	63	61	46	35	56.4%
o4 Mini High	91	81	46	37	24	55.8%
DeepSeek-V2 Chat	100	100	33	30	15	55.5%
Qwen 3.5 35B	100	66	44	35	31	55.3%
Claude Sonnet 4.6	87	69	49	40	29	55.0%
GPT-5	70	69	51	38	35	52.4%
Nemotron 3 Super	100	62	54	33	11	52.0%
Qwen 3.5 122B	89	66	54	37	5	50.1%
Arcee AI: Trinity Mini	74	65	56	30	14	47.8%
Qwen 3.5 9B	100	95	25	16	0	47.3%
Gemini 2.5 Pro	59	51	47	42	33	46.3%
Z.AI GLM 4.7 Flash	65	64	58	20	3	42.1%
Inception Mercury	81	74	34	6	0	39.0%
Gemma 3 4B	61	56	30	20	20	37.3%
ByteDance Seed 2.0 Lite	60	58	36	14	8	35.2%
Gemini 3.1 Flash Lite (Preview)	47	42	31	30	0	29.8%
Mistral NeMO	62	16	7	4	0	17.9%
ByteDance Seed 2.0 Mini	42	23	22	0	0	17.3%
Mistral Small 3.2 24B	52	20	10	0	0	16.5%
ByteDance Seed 1.6	7	0	0	0	0	1.4%

▼

Literary fiction: old friends reunite

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
GPT-5.1	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Claude Sonnet 4.6	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
GPT-5.2	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
MiniMax M2.5	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Claude 3.5 Haiku	100	100	100	100	97	99.4%
Llama 3.1 8B	100	100	100	100	93	98.5%
GPT-5.4 (Reasoning)	100	100	100	100	75	95.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	96	78	94.8%
Claude 3.5 Sonnet	100	100	100	87	87	94.7%
Gemini 3.1 Pro (Preview)	100	100	100	90	80	93.9%
Claude Sonnet 4.5	100	100	100	100	67	93.4%
Claude Opus 4	100	100	100	100	65	93.0%
Mistral Small 4 (Reasoning)	100	100	92	88	72	90.4%
GPT-5.4 Nano (Reasoning, Low)	100	96	93	79	74	88.4%
Llama 3.1 70B	100	100	100	100	35	87.1%
Z.AI GLM 5 Turbo	100	100	100	100	34	86.9%
Claude Opus 4.6	100	100	99	76	57	86.4%
Z.AI GLM 5	100	100	100	100	22	84.5%
Claude 3.7 Sonnet	100	100	91	83	49	84.5%
Claude Opus 4.6 (Reasoning)	100	100	100	78	39	83.4%
GPT-4o, Aug. 6th (temp=1)	100	100	100	63	47	82.0%
Qwen 2.5 72B	100	100	100	100	0	80.0%
Claude Sonnet 4.6 (Reasoning)	100	100	100	62	22	76.9%
Mistral Small 3.2 24B	100	100	100	80	0	75.9%
Ministral 3 14B	100	93	68	62	55	75.7%
Mistral Medium 3.1	100	75	74	65	59	74.6%
GPT-5 Nano	100	96	63	63	51	74.5%
Z.AI GLM 4.5	100	100	67	59	44	73.9%
GPT-5.4 Nano (Reasoning)	79	79	76	72	62	73.6%
Qwen 3.5 Plus (2026-02-15)	100	100	87	60	14	72.2%
GPT-5.4	91	72	69	68	59	71.6%
Grok 4.1 Fast	100	100	94	31	30	70.9%
Writer: Palmyra X5	94	87	74	52	47	70.7%
Mistral Large 2	95	94	67	52	40	69.6%
DeepSeek V3 (2025-03-24)	100	100	100	47	0	69.3%
Grok 4.20 (Beta)	93	78	70	68	37	69.1%
MiniMax M2.7	100	100	100	33	12	68.9%
Claude Sonnet 4	100	100	65	50	20	67.0%
Ministral 3 3B	100	82	75	57	18	66.5%
GPT-5.4 Nano	100	74	70	47	41	66.3%
Rocinante 12B	97	71	64	59	33	64.8%
Gemini 3 Pro (Preview)	100	92	91	32	0	63.0%
Mistral Small 4	90	83	56	45	37	62.0%
MoonshotAI: Kimi K2.5	100	63	54	52	38	61.3%
Hermes 3 70B	83	70	68	45	40	61.2%
GPT-4.1	100	90	48	44	20	60.3%
Arcee AI: Trinity Large (Preview)	100	84	78	26	11	59.8%
Ministral 3 8B	94	91	68	45	0	59.6%
Qwen3 235B A22B Instruct 2507	100	70	61	49	13	58.7%
Mistral Small Creative	87	81	69	55	0	58.3%
Claude Haiku 4.5	100	100	45	42	0	57.4%
Ministral 3B	100	100	52	18	13	56.5%
Grok 4 Fast	75	72	53	48	28	55.2%
Claude 3 Haiku	100	85	51	29	8	54.6%
ByteDance Seed 1.6 Flash	96	72	60	32	12	54.5%
Nemotron 3 Super	100	62	51	46	9	53.7%
Gemini 3 Flash (Preview, Reasoning)	80	53	45	42	41	52.2%
Mistral Large 3	100	91	52	11	0	50.9%
Hermes 3 405B	95	63	62	33	0	50.6%
Mistral NeMO	100	92	42	17	0	50.2%
Gemma 3 12B	100	100	23	13	7	48.5%
o4 Mini High	100	100	40	0	0	48.0%
Gemini 2.5 Flash (Reasoning)	100	80	36	20	0	47.2%
DeepSeek-V2 Chat	100	100	33	0	0	46.7%
Gemini 2.5 Flash	100	96	37	0	0	46.7%
Mistral Large	100	100	32	0	0	46.4%
Gemini 3.1 Flash Lite (Preview)	89	81	60	1	0	46.2%
GPT-4.1 Mini	89	57	30	27	12	43.1%
GPT-4o, May 13th (temp=1)	100	46	36	26	1	41.8%
Gemini 2.5 Flash Lite	100	83	25	0	0	41.5%
LFM2 24B	76	63	42	15	0	39.2%
Z.AI GLM 4.7	100	52	18	17	5	38.4%
GPT-5	100	87	4	0	0	38.1%
GPT-5.4 Mini	65	40	34	31	20	37.9%
GPT-4o, Aug. 6th (temp=0)	71	45	41	33	0	37.7%
Qwen 3 32B	70	36	36	30	10	36.5%
Gemini 3 Flash (Preview)	46	44	39	35	17	36.2%
Cohere Command R+ (Aug. 2024)	82	76	15	3	0	34.9%
ByteDance Seed 2.0 Mini	63	54	53	1	0	34.5%
WizardLM 2 8x22b	98	50	24	0	0	34.4%
GPT-4o Mini (temp=1)	75	49	34	8	0	33.2%
Z.AI GLM 4.7 Flash	80	48	18	11	0	31.3%
Stealth: Aurora Alpha	100	35	20	0	0	31.0%
Qwen 3.5 27B	57	51	41	6	0	30.9%
Inception Mercury	100	28	8	0	0	27.1%
Nemotron 3 Nano	63	47	25	0	0	27.0%
Ministral 8B	100	22	11	0	0	26.6%
ByteDance Seed 2.0 Lite	41	29	29	27	0	25.1%
Qwen 3.5 9B	58	31	28	0	0	23.5%
DeepSeek V3 (2024-12-26)	67	26	20	0	0	22.5%
Aion 2.0	82	30	0	0	0	22.4%
GPT-4o, May 13th (temp=0)	100	7	0	0	0	21.5%
Gemini 2.5 Flash Lite (Reasoning)	60	36	10	0	0	21.2%
Qwen 3.5 Flash	63	39	3	0	0	20.9%
Gemma 3 27B	45	30	30	0	0	20.9%
Stealth: Hunter Alpha	100	2	0	0	0	20.4%
o4 Mini	100	0	0	0	0	20.0%
Qwen 3.5 35B	68	24	0	0	0	18.3%
ByteDance Seed 1.6	65	24	0	0	0	18.0%
Qwen 3.5 397B A17B	28	18	14	11	2	14.6%
Qwen 3.5 122B	49	9	0	0	0	11.7%
Z.AI GLM 4.6	30	28	0	0	0	11.7%
Inception Mercury 2	57	0	0	0	0	11.4%
Stealth: Healer Alpha	46	0	0	0	0	9.2%
GPT-5 Mini	46	0	0	0	0	9.2%
Grok 4	35	0	0	0	0	7.0%
Arcee AI: Trinity Mini	33	0	0	0	0	6.6%
DeepSeek V3.2	27	0	0	0	0	5.5%
Gemini 2.5 Pro	20	7	0	0	0	5.3%
GPT-4o Mini (temp=0)	16	4	0	0	0	4.1%
DeepSeek V3.1	0	0	0	0	0	0.0%
GPT-4.1 Nano	0	0	0	0	0	0.0%
Gemma 3 4B	0	0	0	0	0	0.0%

▼

Mystery: examining a crime scene

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.6 (Reasoning)	100	100	100	100	100	100.0%
Z.AI GLM 5 Turbo	100	100	100	100	100	100.0%
Claude Sonnet 4.6 (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning)	100	100	100	100	100	100.0%
Grok 4.20 (Beta, Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 (Reasoning, Low)	100	100	100	100	100	100.0%
Z.AI GLM 5	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning)	100	100	100	100	100	100.0%
GPT-5.2	100	100	100	100	100	100.0%
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Grok 4	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Grok 4 Fast	100	100	100	100	100	100.0%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Claude 3.5 Sonnet	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Inception Mercury 2	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
GPT-5 Nano	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning)	100	100	100	100	100	100.0%
GPT-5.4 Nano (Reasoning, Low)	100	100	100	100	100	100.0%
Llama 3.1 70B	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Ministral 3 14B	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Ministral 3 3B	100	100	100	100	100	100.0%
Ministral 8B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
Ministral 3B	100	100	100	100	100	100.0%
o4 Mini	100	100	100	100	100	99.9%
Gemma 3 12B	100	100	100	100	100	99.9%
Cohere Command R+ (Aug. 2024)	100	100	100	100	99	99.8%
GPT-4o, May 13th (temp=0)	100	100	100	99	99	99.6%
Claude Sonnet 4.5	100	100	100	100	98	99.6%
DeepSeek V3 (2024-12-26)	100	100	100	100	98	99.6%
Mistral Small 4 (Reasoning)	100	100	100	100	98	99.6%
Qwen3 235B A22B Instruct 2507	100	100	100	100	97	99.4%
Mistral Large 2	100	100	100	100	96	99.3%
GPT-5.1	100	100	100	98	98	99.2%
Claude Opus 4.6	100	100	100	100	94	98.9%
Gemma 3 27B	100	100	100	100	93	98.5%
Gemini 3.1 Pro (Preview)	100	100	100	100	93	98.5%
Nemotron 3 Nano	100	100	100	100	90	98.1%
LFM2 24B	100	100	100	100	89	97.9%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	88	97.6%
Stealth: Aurora Alpha	100	100	100	100	88	97.6%
GPT-4o, May 13th (temp=1)	100	100	100	100	87	97.3%
GPT-5 Mini	100	100	100	94	92	97.3%
ByteDance Seed 1.6 Flash	100	100	100	100	85	97.0%
Gemini 3 Pro (Preview)	100	100	100	100	85	97.0%
GPT-5.4 Nano	100	100	100	98	85	96.7%
Ministral 3 8B	100	100	100	98	84	96.5%
Claude 3.5 Haiku	100	100	100	95	87	96.3%
Rocinante 12B	100	100	100	97	82	95.9%
Mistral Small Creative	100	100	100	91	87	95.5%
Nemotron 3 Super	100	100	100	100	76	95.3%
MoonshotAI: Kimi K2.5	100	100	100	100	71	94.3%
GPT-4o, Aug. 6th (temp=0)	100	100	100	99	71	94.0%
GPT-4o Mini (temp=0)	100	100	100	100	68	93.7%
MiniMax M2.5	100	100	100	100	66	93.1%
Qwen 2.5 72B	100	100	98	92	76	93.1%
Mistral Large 3	100	100	100	97	68	93.0%
Hermes 3 405B	100	100	100	100	64	92.9%
Arcee AI: Trinity Large (Preview)	100	100	100	90	73	92.5%
Mistral Large	100	100	100	93	67	92.0%
Qwen 3 32B	100	100	100	90	70	91.9%
Claude Haiku 4.5	100	100	100	99	59	91.5%
Z.AI GLM 4.7 Flash	100	100	93	84	79	91.3%
Mistral Medium 3.1	100	100	96	88	72	91.1%
Writer: Palmyra X5	100	100	100	100	55	91.1%
Gemini 3 Flash (Preview)	100	100	95	78	75	89.4%
GPT-4.1 Mini	100	100	100	82	62	88.8%
Gemini 2.5 Flash	100	100	85	80	79	88.7%
Aion 2.0	100	100	100	83	60	88.7%
GPT-4o Mini (temp=1)	100	100	100	84	58	88.3%
Qwen 3.5 397B A17B	100	100	84	78	76	87.5%
MiniMax M2.7	100	97	85	83	72	87.2%
Z.AI GLM 4.7	100	100	100	87	42	85.7%
Hermes 3 70B	100	100	93	76	54	84.6%
Claude Sonnet 4.6	100	100	84	79	59	84.4%
GPT-5	99	89	80	77	76	84.3%
DeepSeek-V2 Chat	100	100	100	87	34	84.1%
GPT-4.1	100	100	100	77	42	83.7%
Stealth: Healer Alpha	90	90	87	87	64	83.5%
Inception Mercury	100	100	78	75	55	81.7%
ByteDance Seed 2.0 Mini	100	93	79	74	59	81.0%
Gemini 2.5 Flash Lite (Reasoning)	89	82	79	78	72	80.0%
WizardLM 2 8x22b	100	100	87	82	27	79.1%
DeepSeek V3.2	100	85	76	68	65	79.1%
ByteDance Seed 2.0 Lite	100	100	85	62	48	79.1%
Gemini 2.5 Flash (Reasoning)	100	85	83	63	61	78.5%
Qwen 3.5 Flash	100	92	89	59	52	78.3%
Qwen 3.5 27B	100	100	97	70	24	78.2%
Stealth: Hunter Alpha	98	82	80	65	61	77.1%
GPT-4o, Aug. 6th (temp=1)	100	100	89	49	47	77.1%
Gemini 2.5 Flash Lite	100	100	81	68	36	77.0%
Qwen 3.5 122B	100	85	75	71	53	76.8%
Gemma 3 4B	100	97	94	62	31	76.8%
Qwen 3.5 9B	100	95	87	86	15	76.4%
Gemini 3 Flash (Preview, Reasoning)	87	77	73	70	69	75.0%
Z.AI GLM 4.6	100	79	77	63	53	74.4%
Gemini 2.5 Pro	94	90	68	67	50	73.8%
o4 Mini High	100	75	73	71	45	72.7%
DeepSeek V3.1	100	80	60	48	45	66.6%
Qwen 3.5 35B	94	87	74	71	5	66.1%
Arcee AI: Trinity Mini	79	71	71	62	33	63.4%
GPT-4.1 Nano	81	74	61	60	36	62.6%
Gemini 3.1 Flash Lite (Preview)	59	53	52	49	37	49.8%
ByteDance Seed 1.6	71	54	49	35	34	48.6%
Mistral NeMO	51	50	41	26	16	36.8%
Mistral Small 3.2 24B	100	38	0	0	0	27.7%

▼

Romance: separated couple reunites

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Ministral 3B	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	99	95	98.8%
Llama 3.1 Nemotron 70B	100	100	100	100	84	96.8%
Z.AI GLM 5 Turbo	100	99	95	89	83	93.2%
GPT-5.2	100	100	100	100	62	92.3%
Gemini 3.1 Pro (Preview)	97	95	89	87	82	89.9%
Mistral Large 3	100	100	100	87	58	89.0%
Claude 3.5 Haiku	100	100	87	80	77	88.8%
Ministral 3 3B	100	100	100	100	13	82.6%
GPT-5.4	100	100	74	73	64	82.4%
GPT-5.4 Nano (Reasoning)	94	85	83	77	69	81.5%
GPT-5.4 Nano (Reasoning, Low)	100	100	77	67	60	81.0%
Mistral Medium 3.1	100	100	100	85	13	79.7%
Mistral Large 2	100	100	76	61	58	79.0%
GPT-5.4 Nano	100	82	74	72	66	78.7%
Mistral Small 4 (Reasoning)	88	85	83	80	55	78.5%
Llama 3.1 70B	100	100	95	84	10	77.9%
GPT-5.4 Mini (Reasoning)	100	82	80	72	53	77.4%
Mistral Small 4	100	100	85	71	31	77.2%
Z.AI GLM 4.5	100	100	76	57	52	76.9%
GPT-5.4 Mini	100	81	79	75	43	75.4%
Claude 3 Haiku	83	80	77	69	61	74.2%
Ministral 3 14B	89	88	73	68	51	74.0%
Claude Sonnet 4	100	82	75	75	24	71.1%
Ministral 8B	98	81	62	55	54	69.9%
GPT-5.4 Mini (Reasoning, Low)	85	83	58	55	53	66.8%
Claude 3.7 Sonnet	100	100	75	48	7	66.0%
Rocinante 12B	100	100	80	41	0	64.2%
GPT-5 Nano	100	93	85	40	0	63.8%
Claude Haiku 4.5	100	100	47	36	33	63.2%
GPT-5.4 (Reasoning, Low)	100	61	57	50	48	63.1%
Mistral Small Creative	92	83	69	37	31	62.6%
Grok 4.1 Fast	100	56	55	53	49	62.5%
Grok 4.20 (Beta)	86	71	62	54	38	62.0%
Grok 4.20 (Beta, Reasoning)	76	73	56	52	48	60.9%
Hermes 3 405B	100	100	58	45	2	60.8%
Ministral 3 8B	100	79	69	37	8	58.6%
Z.AI GLM 5	80	63	53	50	40	57.3%
Arcee AI: Trinity Large (Preview)	75	71	70	44	25	56.9%
Qwen 3 32B	100	71	65	47	0	56.5%
Gemini 3 Flash (Preview)	87	81	46	37	23	54.7%
Hermes 3 70B	100	76	70	20	0	53.1%
Inception Mercury 2	77	65	62	55	0	51.8%
DeepSeek V3 (2024-12-26)	100	75	74	0	0	49.8%
GPT-5.4 (Reasoning)	67	66	47	41	29	49.7%
GPT-4o, Aug. 6th (temp=1)	89	74	69	4	1	47.3%
Claude Sonnet 4.5	100	66	66	2	0	46.8%
GPT-4o, Aug. 6th (temp=0)	96	68	59	9	0	46.6%
Claude Opus 4.5	63	55	43	36	34	46.2%
Claude Opus 4.6 (Reasoning)	58	53	50	45	22	45.9%
Qwen 3.5 Plus (2026-02-15)	72	43	41	40	29	44.9%
Claude Opus 4.6	71	54	52	22	13	42.2%
Mistral Large	70	68	52	20	0	41.9%
Gemini 3 Flash (Preview, Reasoning)	64	57	41	37	7	41.3%
GPT-4o, May 13th (temp=1)	78	64	47	16	0	41.0%
Qwen 2.5 72B	100	52	37	9	6	40.9%
Writer: Palmyra X5	100	61	22	17	0	40.1%
DeepSeek V3 (2025-03-24)	60	54	33	27	23	39.5%
Grok 4 Fast	58	56	45	20	12	38.2%
Mistral Small 3.2 24B	99	43	40	9	0	38.2%
Claude Opus 4	53	48	41	38	0	35.9%
MoonshotAI: Kimi K2.5	54	49	48	24	2	35.3%
Qwen3 235B A22B Instruct 2507	100	38	27	12	0	35.2%
Z.AI GLM 4.6	91	34	18	15	12	34.2%
Gemma 3 12B	59	40	39	23	6	33.2%
GPT-4.1 Mini	68	51	40	0	0	31.9%
GPT-5.1	53	33	27	21	8	28.5%
Nemotron 3 Super	57	56	15	11	0	27.9%
Qwen 3.5 397B A17B	50	33	19	14	12	25.7%
Claude Sonnet 4.6 (Reasoning)	41	28	27	20	12	25.4%
ByteDance Seed 1.6 Flash	47	46	27	6	0	25.3%
LFM2 24B	52	39	29	0	0	24.2%
Cohere Command R+ (Aug. 2024)	84	23	7	0	0	22.7%
Claude 3.5 Sonnet	34	26	20	20	5	20.9%
DeepSeek V3.2	55	37	12	0	0	20.8%
Qwen 3.5 35B	42	34	27	0	0	20.7%
Qwen 3.5 9B	49	45	0	0	0	18.8%
ByteDance Seed 1.6	73	20	0	0	0	18.7%
Gemma 3 27B	30	27	17	15	0	17.8%
Z.AI GLM 4.7	39	24	15	9	0	17.5%
Grok 4	43	32	10	1	0	17.1%
DeepSeek-V2 Chat	64	14	8	0	0	17.1%
MiniMax M2.7	43	42	0	0	0	17.0%
GPT-4.1	69	14	0	0	0	16.6%
ByteDance Seed 2.0 Lite	46	20	15	0	0	16.2%
Stealth: Healer Alpha	51	16	14	0	0	16.1%
Mistral NeMO	48	32	0	0	0	16.0%
Qwen 3.5 27B	56	12	4	0	0	14.5%
Gemini 3 Pro (Preview)	38	20	12	0	0	14.1%
Stealth: Hunter Alpha	32	24	12	0	0	13.7%
Arcee AI: Trinity Mini	43	12	0	0	0	11.1%
Stealth: Aurora Alpha	22	20	13	0	0	10.9%
o4 Mini	53	0	0	0	0	10.5%
Gemini 2.5 Flash (Reasoning)	39	12	0	0	0	10.4%
Claude Sonnet 4.6	24	11	7	2	0	8.7%
GPT-4o Mini (temp=1)	20	13	10	0	0	8.7%
WizardLM 2 8x22b	34	4	0	0	0	7.7%
MiniMax M2.5	28	6	3	0	0	7.4%
Qwen 3.5 Flash	27	9	0	0	0	7.2%
ByteDance Seed 2.0 Mini	17	14	0	0	0	6.3%
Gemma 3 4B	17	14	0	0	0	6.2%
o4 Mini High	20	7	4	0	0	6.2%
Gemini 2.5 Flash	23	7	0	0	0	6.0%
GPT-5 Mini	28	0	0	0	0	5.6%
DeepSeek V3.1	28	0	0	0	0	5.5%
Inception Mercury	27	0	0	0	0	5.4%
Gemini 3.1 Flash Lite (Preview)	26	0	0	0	0	5.1%
Z.AI GLM 4.7 Flash	23	2	0	0	0	5.1%
GPT-4o, May 13th (temp=0)	25	0	0	0	0	5.1%
Aion 2.0	15	5	0	0	0	4.2%
Qwen 3.5 122B	15	2	0	0	0	3.5%
Gemini 2.5 Flash Lite	17	0	0	0	0	3.5%
GPT-4.1 Nano	17	0	0	0	0	3.4%
Nemotron 3 Nano	16	0	0	0	0	3.3%
Gemini 2.5 Pro	4	0	0	0	0	0.8%
Gemini 2.5 Flash Lite (Reasoning)	4	0	0	0	0	0.7%
GPT-5	0	0	0	0	0	0.0%
GPT-4o Mini (temp=0)	0	0	0	0	0	0.0%

▼

Thriller: chase through city streets

Model	# 1	# 2	# 3	# 4	# 5	Avg ▼
Claude Opus 4.5	100	100	100	100	100	100.0%
Grok 4.1 Fast	100	100	100	100	100	100.0%
Claude Sonnet 4	100	100	100	100	100	100.0%
Claude Sonnet 4.5	100	100	100	100	100	100.0%
Claude Opus 4	100	100	100	100	100	100.0%
Z.AI GLM 4.5	100	100	100	100	100	100.0%
Claude Haiku 4.5	100	100	100	100	100	100.0%
GPT-5.4	100	100	100	100	100	100.0%
Grok 4.20 (Beta)	100	100	100	100	100	100.0%
Claude 3.7 Sonnet	100	100	100	100	100	100.0%
Hermes 3 405B	100	100	100	100	100	100.0%
GPT-5.4 Mini	100	100	100	100	100	100.0%
Mistral Large 2	100	100	100	100	100	100.0%
Mistral Small 4 (Reasoning)	100	100	100	100	100	100.0%
DeepSeek V3 (2025-03-24)	100	100	100	100	100	100.0%
Mistral Medium 3.1	100	100	100	100	100	100.0%
Llama 3.1 Nemotron 70B	100	100	100	100	100	100.0%
Claude 3 Haiku	100	100	100	100	100	100.0%
Llama 3.1 8B	100	100	100	100	100	100.0%
LFM2 24B	100	100	100	100	100	100.0%
Mistral Small 4	100	100	100	100	99	99.8%
Mistral Large 3	100	100	100	100	98	99.7%
Grok 4 Fast	100	100	100	100	98	99.6%
Ministral 3 14B	100	100	100	100	97	99.5%
GPT-5.4 (Reasoning, Low)	100	100	100	100	95	98.9%
Ministral 3 3B	100	100	100	100	93	98.7%
GPT-4.1 Mini	100	100	100	98	93	98.2%
GPT-5.4 (Reasoning)	100	100	100	100	91	98.1%
Grok 4.20 (Beta, Reasoning)	100	100	100	96	94	97.9%
Z.AI GLM 5 Turbo	100	100	100	96	93	97.9%
GPT-5.2	100	100	100	100	88	97.6%
Qwen 3.5 Plus (2026-02-15)	100	100	100	100	88	97.6%
GPT-5.4 Mini (Reasoning)	100	100	100	98	89	97.5%
Stealth: Hunter Alpha	100	100	100	93	92	97.1%
Qwen 3 32B	100	100	96	95	93	96.6%
Claude 3.5 Sonnet	100	100	100	99	83	96.4%
Claude Opus 4.6 (Reasoning)	100	100	100	100	82	96.3%
Claude 3.5 Haiku	100	100	100	100	80	96.0%
Ministral 3B	100	100	100	98	82	96.0%
Qwen3 235B A22B Instruct 2507	100	100	100	100	78	95.6%
Mistral Large	100	100	100	97	79	95.4%
Hermes 3 70B	100	100	97	90	89	95.3%
Cohere Command R+ (Aug. 2024)	100	100	100	100	76	95.1%
GPT-5.4 Mini (Reasoning, Low)	100	100	100	90	83	94.6%
MiniMax M2.5	100	100	100	93	79	94.3%
GPT-4o, May 13th (temp=1)	100	100	99	87	85	94.0%
Gemini 3.1 Pro (Preview)	100	100	100	100	68	93.6%
Arcee AI: Trinity Mini	100	100	100	84	81	93.0%
Z.AI GLM 5	100	100	100	92	70	92.4%
GPT-5 Nano	100	100	100	84	78	92.3%
Llama 3.1 70B	100	100	100	100	62	92.3%
Mistral Small Creative	100	100	100	94	67	92.1%
Claude Opus 4.6	100	100	100	100	55	91.0%
MiniMax M2.7	100	100	100	88	62	90.1%
Grok 4	100	100	98	82	67	89.4%
Z.AI GLM 4.6	100	100	100	90	57	89.3%
Arcee AI: Trinity Large (Preview)	100	100	91	83	71	89.0%
Gemini 2.5 Flash Lite	100	100	100	87	58	89.0%
GPT-4o, Aug. 6th (temp=0)	100	100	88	78	77	88.6%
Ministral 3 8B	100	100	95	90	56	88.2%
Writer: Palmyra X5	100	100	92	85	60	87.5%
DeepSeek-V2 Chat	100	100	100	68	67	86.9%
Gemini 2.5 Flash	100	100	83	74	73	86.2%
Gemini 2.5 Flash Lite (Reasoning)	100	100	88	77	63	85.7%
Stealth: Aurora Alpha	100	100	100	75	53	85.6%
Gemma 3 12B	100	100	83	80	66	85.6%
GPT-5.4 Nano (Reasoning, Low)	100	88	87	79	74	85.5%
MoonshotAI: Kimi K2.5	100	100	100	75	52	85.4%
GPT-4.1	100	100	100	92	33	85.1%
GPT-5.1	100	93	87	84	61	84.9%
GPT-4o, May 13th (temp=0)	100	100	76	73	72	84.2%
GPT-4o, Aug. 6th (temp=1)	100	100	89	81	45	82.8%
Ministral 8B	100	100	100	59	50	81.8%
Rocinante 12B	100	100	100	79	25	80.8%
Qwen 2.5 72B	100	100	94	60	49	80.6%
Nemotron 3 Super	100	89	82	77	47	79.0%
Gemini 3 Flash (Preview)	96	90	90	71	49	78.9%
Gemma 3 27B	87	85	78	75	69	78.8%
Z.AI GLM 4.7	88	87	86	82	50	78.3%
Nemotron 3 Nano	100	100	70	67	54	78.2%
Gemini 3 Pro (Preview)	98	95	90	55	48	77.1%
Gemini 2.5 Flash (Reasoning)	90	88	84	77	47	77.0%
Aion 2.0	100	84	78	72	51	77.0%
GPT-5.4 Nano	89	86	80	65	62	76.5%
GPT-4o Mini (temp=1)	100	100	81	79	22	76.4%
Stealth: Healer Alpha	95	93	79	60	50	75.3%
Gemma 3 4B	100	100	90	46	39	75.1%
GPT-5.4 Nano (Reasoning)	91	85	74	65	59	74.6%
Gemini 3 Flash (Preview, Reasoning)	84	82	78	68	60	74.3%
o4 Mini High	100	92	88	60	25	72.8%
Inception Mercury 2	100	100	99	64	1	72.8%
GPT-4.1 Nano	100	85	78	54	44	72.2%
ByteDance Seed 1.6 Flash	84	83	66	66	54	70.7%
o4 Mini	100	94	76	74	8	70.1%
DeepSeek V3.2	75	74	71	67	61	69.6%
DeepSeek V3.1	84	70	67	63	39	64.6%
Claude Sonnet 4.6	100	80	79	30	30	63.9%
WizardLM 2 8x22b	92	73	69	44	33	62.2%
Qwen 3.5 397B A17B	92	89	83	32	0	59.1%
Qwen 3.5 Flash	82	72	62	45	29	58.1%
Qwen 3.5 9B	94	84	40	39	32	57.8%
Qwen 3.5 27B	75	59	58	49	40	56.5%
Mistral Small 3.2 24B	78	78	70	41	14	56.0%
Gemini 3.1 Flash Lite (Preview)	84	77	45	44	30	55.9%
Gemini 2.5 Pro	65	62	50	46	36	52.1%
ByteDance Seed 2.0 Lite	70	70	58	58	4	52.0%
ByteDance Seed 2.0 Mini	64	53	52	51	36	51.0%
DeepSeek V3 (2024-12-26)	100	48	39	32	23	48.3%
GPT-5	77	57	57	44	4	47.9%
Z.AI GLM 4.7 Flash	91	59	51	38	0	47.7%
GPT-5 Mini	76	54	42	32	27	46.3%
Qwen 3.5 122B	79	59	43	23	22	45.0%
Claude Sonnet 4.6 (Reasoning)	91	56	44	24	6	44.1%
GPT-4o Mini (temp=0)	53	52	40	34	33	42.3%
ByteDance Seed 1.6	78	71	29	14	13	40.8%
Qwen 3.5 35B	66	44	44	40	0	38.5%
Mistral NeMO	59	45	43	11	9	33.2%
Inception Mercury	23	0	0	0	0	4.5%

Pronoun-first sentence starts

Overall Performance

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)

Individual Scenarios

Detailed Writing Rules

Fantasy: entering an ancient ruin

Horror: alone in an eerie place at night

Literary fiction: old friends reunite

Mystery: examining a crime scene

Romance: separated couple reunites

Thriller: chase through city streets

genre

Fantasy: entering an ancient ruin

Horror: alone in an eerie place at night

Literary fiction: old friends reunite

Mystery: examining a crime scene

Romance: separated couple reunites

Thriller: chase through city streets

Novelcrafter Default Prompt

Fantasy: entering an ancient ruin

Horror: alone in an eerie place at night

Literary fiction: old friends reunite

Mystery: examining a crime scene

Romance: separated couple reunites

Thriller: chase through city streets