Fantasy: entering an ancient ruin

Bad Writing Habits

Detects common prose quality anti-patterns in AI-generated creative writing, including passive voice, past progressive overuse, weak dialogue tags, filter words, purple prose, cliches, AI-ism words/adverbs/names, and more.

Creative Writing Hallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4 (Reasoning)	92%
GPT-5.4	89%
GPT-5.4 (Reasoning, Low)	89%
GPT-5.4 Mini (Reasoning, Low)	87%
GPT-5.4 Mini (Reasoning)	87%
Claude Sonnet 4.5	87%
Claude Sonnet 4.6	87%
Claude Sonnet 4.6 (Reasoning)	87%
Claude Opus 4.6	86%
GPT-5.1	86%
GPT-5.4 Mini	86%
GPT-5	85%
Z.AI GLM 5 Turbo	85%
Claude Opus 4.6 (Reasoning)	85%
Qwen 3.5 397B A17B	84%
Qwen3 235B A22B Instruct 2507	84%
Claude Opus 4	84%
Writer: Palmyra X5	84%
Gemini 3.1 Pro (Preview)	83%
Claude Opus 4.5	83%

	Score	Cost	Time
GPT-5.4 Mini (Reasoning)	87%	$0.018	22.3s
GPT-5.4 Mini (Reasoning, Low)	87%	$0.014	16.2s
GPT-5.4 Mini	86%	$0.016	17.1s
GPT-5.4	89%	$0.039	1.2m
Z.AI GLM 5 Turbo	85%	$0.0090	25.9s
GPT-5.4 (Reasoning, Low)	89%	$0.056	1.3m
Claude Sonnet 4.6	87%	$0.040	37.2s
Writer: Palmyra X5	84%	$0.013	22.1s
Claude Sonnet 4.5	87%	$0.045	39.5s
Qwen 3.5 35B	81%	$0.043	2.3m
Z.AI GLM 5	83%	$0.0095	1.5m
DeepSeek V3 (2024-12-26)	78%	$0.0029	35.6s
MiniMax M2.7	82%	$0.0032	59.3s
Qwen 3.5 Flash	81%	$0.0033	45.4s
Qwen 3.5 122B	80%	$0.016	39.5s
Qwen3 235B A22B Instruct 2507	84%	$0.0017	1.1m
GPT-5.4 (Reasoning)	92%	$0.081	2.3m
Qwen 3.5 397B A17B	84%	$0.0048	3.3m
GPT-4.1	79%	$0.021	40.8s
DeepSeek-V2 Chat	78%	$0.0029	42.0s

	Score	Cost	Speed	Stability
GPT-5.4 Mini (Reasoning, Low)	87%	$0.014	16.2s	84%
GPT-5.4	89%	$0.039	1.2m	88%
GPT-5.4 Mini (Reasoning)	87%	$0.018	22.3s	81%
GPT-5.4 (Reasoning)	92%	$0.081	2.3m	90%
GPT-5.4 Mini	86%	$0.016	17.1s	81%
Claude Sonnet 4.6	87%	$0.040	37.2s	84%
Claude Sonnet 4.5	87%	$0.045	39.5s	83%
Z.AI GLM 5 Turbo	85%	$0.0090	25.9s	80%
GPT-5.4 (Reasoning, Low)	89%	$0.056	1.3m	84%
GPT-5.1	86%	$0.053	1.8m	85%
Writer: Palmyra X5	84%	$0.013	22.1s	76%
Claude Opus 4.6	86%	$0.087	1.1m	82%
Qwen 3.5 Flash	81%	$0.0033	45.4s	76%
Qwen 3.5 122B	80%	$0.016	39.5s	77%
Grok 4.20 (Beta, Reasoning)	81%	$0.030	24.6s	77%
Qwen 3.5 9B	81%	$0.0011	43.6s	75%
Qwen3 235B A22B Instruct 2507	84%	$0.0017	1.1m	73%
GPT-5.4 Nano (Reasoning)	79%	$0.0060	26.5s	76%
MiniMax M2.7	82%	$0.0032	59.3s	74%
o4 Mini	79%	$0.019	34.0s	77%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
4	GPT-5.4 (Reasoning)	$0.081	2.3m	90%	94	93	92	92	90	92%
2	GPT-5.4	$0.039	1.2m	88%	91	91	90	88	87	89%
9	GPT-5.4 (Reasoning, Low)	$0.056	1.3m	84%	93	92	91	86	83	89%
1	GPT-5.4 Mini (Reasoning, Low)	$0.014	16.2s	84%	90	89	87	86	85	87%
3	GPT-5.4 Mini (Reasoning)	$0.018	22.3s	81%	92	90	88	84	82	87%
7	Claude Sonnet 4.5	$0.045	39.5s	83%	91	88	87	87	82	87%
6	Claude Sonnet 4.6	$0.040	37.2s	84%	89	89	87	86	84	87%
40	Claude Sonnet 4.6 (Reasoning)	$0.139	2.3m	82%	91	87	86	85	84	87%
12	Claude Opus 4.6	$0.087	1.1m	82%	89	88	86	85	84	86%
10	GPT-5.1	$0.053	1.8m	85%	87	87	86	85	85	86%
5	GPT-5.4 Mini	$0.016	17.1s	81%	90	89	87	84	81	86%
37	GPT-5	$0.064	3.0m	80%	88	87	84	83	82	85%
8	Z.AI GLM 5 Turbo	$0.0090	25.9s	80%	88	87	86	83	79	85%
22	Claude Opus 4.6 (Reasoning)	$0.103	1.4m	82%	87	85	84	84	83	85%
21	Qwen 3.5 397B A17B	$0.0048	3.3m	81%	87	86	86	83	79	84%
17	Qwen3 235B A22B Instruct 2507	$0.0017	1.1m	73%	90	90	82	81	77	84%
116	Claude Opus 4	$0.305	1.9m	76%	91	87	85	79	77	84%
11	Writer: Palmyra X5	$0.013	22.1s	76%	90	87	83	80	79	84%
72	Gemini 3.1 Pro (Preview)	$0.136	2.2m	78%	88	85	84	80	79	83%
28	Claude Opus 4.5	$0.079	48.4s	77%	86	85	85	85	74	83%
27	Z.AI GLM 5	$0.0095	1.5m	74%	89	87	84	80	73	83%
19	MiniMax M2.7	$0.0032	59.3s	74%	90	85	81	78	77	82%
39	Qwen 3.5 35B	$0.043	2.3m	78%	84	82	82	81	77	81%
13	Qwen 3.5 Flash	$0.0033	45.4s	76%	86	81	81	79	78	81%
16	Qwen 3.5 9B	$0.0011	43.6s	75%	85	82	79	79	79	81%
15	Grok 4.20 (Beta, Reasoning)	$0.030	24.6s	77%	83	83	81	79	77	81%
43	ByteDance Seed 2.0 Lite	$0.011	2.0m	74%	87	83	82	77	73	80%
14	Qwen 3.5 122B	$0.016	39.5s	77%	83	83	81	78	77	80%
26	Grok 4.20 (Beta)	$0.018	14.5s	72%	85	84	78	78	75	80%
23	GPT-4.1	$0.021	40.8s	76%	82	82	82	78	73	79%
29	DeepSeek V3.2	$0.0022	1.1m	74%	85	79	79	79	76	79%
52	Hermes 3 70B	$0.0021	2.0m	72%	85	82	79	78	73	79%
24	Qwen 3.5 27B	$0.012	1.1m	77%	81	80	79	79	77	79%
36	Rocinante 12B	$0.0018	23.4s	69%	89	80	78	77	70	79%
20	o4 Mini	$0.019	34.0s	77%	81	80	79	78	76	79%
18	GPT-5.4 Nano (Reasoning)	$0.0060	26.5s	76%	82	79	79	77	76	79%
60	GPT-5.2	$0.057	1.4m	75%	81	80	78	78	74	78%
54	o4 Mini High	$0.028	53.0s	71%	84	81	77	75	74	78%
35	DeepSeek V3 (2025-03-24)	$0.0017	30.2s	71%	85	77	76	76	75	78%
53	DeepSeek-V2 Chat	$0.0029	42.0s	68%	87	79	79	78	66	78%
47	Mistral Large 2	$0.020	32.2s	71%	86	77	77	75	74	78%
98	ByteDance Seed 2.0 Mini	$0.0046	4.6m	73%	82	80	78	77	73	78%
25	Grok 4.1 Fast	$0.0024	50.4s	76%	79	78	78	78	75	78%
93	MoonshotAI: Kimi K2.5	$0.021	3.3m	72%	83	78	78	77	72	78%
45	Aion 2.0	$0.0088	1.4m	74%	80	80	77	76	75	78%
50	Mistral Large	$0.024	46.8s	72%	82	80	76	76	75	78%
48	Gemini 3 Flash (Preview, Reasoning)	$0.012	26.2s	69%	83	82	77	76	69	78%
31	DeepSeek V3 (2024-12-26)	$0.0029	35.6s	73%	82	81	80	74	71	78%
34	GPT-5.4 Nano (Reasoning, Low)	$0.0051	19.3s	72%	81	81	77	76	72	77%
82	ByteDance Seed 1.6	$0.011	2.0m	70%	83	79	75	75	73	77%
79	Z.AI GLM 4.7	$0.011	1.3m	67%	84	82	79	77	63	77%
32	Gemma 3 27B	$0.0010	45.9s	74%	80	78	77	76	75	77%
66	Hermes 3 405B	$0.0062	49.2s	67%	87	77	76	73	72	77%
84	WizardLM 2 8x22b	$0.0045	3.3m	73%	80	79	76	75	75	77%
33	Gemini 2.5 Flash	$0.0057	10.6s	72%	82	78	78	74	72	77%
55	GPT-5 Mini	$0.010	1.0m	71%	82	77	76	74	74	77%
58	Gemini 3 Pro (Preview)	$0.057	51.6s	74%	78	78	76	76	75	77%
30	GPT-5.4 Nano	$0.0050	18.4s	74%	78	78	76	76	75	77%
68	Gemini 2.5 Pro	$0.039	37.7s	70%	82	80	79	76	65	76%
63	Qwen 3 32B	$0.0020	45.5s	68%	83	83	78	69	68	76%
42	Ministral 3 14B	$0.0013	9.5s	70%	83	79	77	71	70	76%
38	ByteDance Seed 1.6 Flash	$0.0015	30.1s	72%	79	78	75	74	74	76%
41	Mistral Small 4 (Reasoning)	$0.0027	32.6s	72%	80	76	76	75	72	76%
76	Claude 3.5 Sonnet	$0.066	35.6s	72%	79	78	77	72	71	75%
70	Mistral Large 3	$0.0064	51.4s	68%	83	78	75	71	70	75%
64	DeepSeek V3.1	$0.0027	1.6m	72%	79	77	76	74	71	75%
57	Qwen 3.5 Plus (2026-02-15)	$0.0073	37.0s	70%	79	78	75	75	70	75%
46	Grok 4 Fast	$0.0020	18.4s	71%	79	76	75	74	72	75%
44	Claude 3 Haiku	$0.0034	19.4s	72%	78	75	74	74	73	75%
49	Gemma 3 12B	$0.0004	43.8s	72%	77	76	76	75	70	75%
51	Mistral Small 4	$0.0019	19.9s	71%	79	77	75	73	70	75%
78	GPT-4o, May 13th (temp=1)	$0.048	14.9s	70%	78	77	73	73	72	75%
65	GPT-4o, Aug. 6th (temp=1)	$0.022	18.5s	69%	80	78	77	72	66	75%
56	Qwen 2.5 72B	$0.0017	49.7s	72%	76	76	76	75	69	74%
83	Claude Haiku 4.5	$0.015	23.3s	66%	80	77	72	71	69	74%
87	Stealth: Hunter Alpha	$0.0000	43.3s	65%	82	74	71	71	70	74%
94	Grok 4	$0.045	1.3m	70%	78	74	73	73	71	74%
90	Z.AI GLM 4.7 Flash	$0.0019	1.2m	66%	82	76	73	69	69	74%
74	LFM2 24B	$0.0003	28.9s	67%	80	76	73	71	69	74%
81	MiniMax M2.5	$0.0035	52.2s	68%	79	75	74	72	68	74%
67	Z.AI GLM 4.6	$0.0073	31.3s	70%	77	74	72	72	71	73%
61	Stealth: Healer Alpha	$0.0000	26.0s	70%	77	76	76	72	65	73%
75	Gemma 3 4B	$0.0003	22.4s	67%	79	74	71	71	71	73%
80	GPT-4o, Aug. 6th (temp=0)	$0.020	17.0s	68%	78	75	73	73	67	73%
73	Mistral Small Creative	$0.0010	8.9s	67%	78	78	73	70	67	73%
71	GPT-4o Mini (temp=1)	$0.0016	50.4s	70%	76	74	73	72	71	73%
62	GPT-4.1 Mini	$0.0037	26.1s	70%	76	75	74	70	70	73%
88	Mistral Medium 3.1	$0.0061	44.7s	67%	78	77	72	69	68	73%
86	Cohere Command R+ (Aug. 2024)	$0.029	41.0s	69%	76	75	74	71	67	73%
100	Claude Sonnet 4	$0.040	40.5s	66%	79	76	71	71	67	73%
69	Gemini 2.5 Flash (Reasoning)	$0.013	22.2s	70%	76	72	72	72	72	73%
59	Mistral NeMO	$0.0010	10.4s	70%	74	73	73	73	68	72%
77	Gemini 2.5 Flash Lite	$0.0012	8.8s	67%	77	75	72	70	67	72%
105	Llama 3.1 8B	$0.0003	55.5s	61%	81	79	71	65	64	72%
89	Ministral 3 8B	$0.0010	8.2s	64%	77	77	70	68	68	72%
95	GPT-4o Mini (temp=0)	$0.0015	40.6s	65%	76	75	72	69	63	71%
92	Z.AI GLM 4.5	$0.0065	40.9s	67%	74	73	72	72	64	71%
91	Ministral 3B	$0.0002	3.4s	64%	77	74	72	71	60	71%
101	GPT-4o, May 13th (temp=0)	$0.053	18.7s	67%	74	73	71	70	67	71%
85	Gemini 3 Flash (Preview)	$0.0083	19.2s	68%	73	72	70	69	69	71%
99	Arcee AI: Trinity Large (Preview)	$0.0000	1.1m	66%	77	72	71	68	65	71%
102	Ministral 8B	$0.0007	13.2s	62%	79	72	68	66	66	70%
96	Gemini 3.1 Flash Lite (Preview)	$0.0039	8.4s	65%	75	69	68	68	67	69%
115	Claude 3.7 Sonnet	$0.057	55.5s	65%	73	71	71	69	61	69%
97	Claude 3.5 Haiku	$0.0058	10.3s	65%	72	70	68	68	67	69%
110	Llama 3.1 Nemotron 70B	$0.0072	31.4s	60%	76	73	67	65	62	69%
103	Gemini 2.5 Flash Lite (Reasoning)	$0.0033	29.8s	65%	72	70	69	66	64	68%
109	Llama 3.1 70B	$0.0038	28.0s	61%	75	71	69	67	58	68%
106	GPT-4.1 Nano	$0.0010	15.1s	62%	75	68	68	67	63	68%
108	Nemotron 3 Super	$0.0000	43.3s	63%	73	68	68	67	62	68%
112	Nemotron 3 Nano	$0.0011	54.1s	62%	74	71	68	63	62	68%
107	Arcee AI: Trinity Mini	$0.0005	8.0s	62%	72	71	68	65	59	67%
104	Inception Mercury	$0.012	16.1s	66%	68	68	68	65	65	67%
114	Inception Mercury 2	$0.0099	13.7s	60%	71	71	66	63	62	67%
117	GPT-5 Nano	$0.0042	1.3m	62%	69	65	65	64	63	65%
113	Ministral 3 3B	$0.0006	3.5s	61%	69	68	67	62	57	65%
111	Stealth: Aurora Alpha	$0.0000	13.1s	63%	66	66	65	62	62	64%
118	Mistral Small 3.2 24B	$0.0083	7.6m	58%	67	66	63	62	56	63%
76.40%

Median	Evaluator	Top 3	Flop 3
72.3%	"Not X but Y" pattern overuse	100DeepSeek V3 (2024-12-26) 100ByteDance Seed 1.6 100Llama 3.1 8B	0GPT-5 Nano 0Stealth: Hunter Alpha 0Gemini 2.5 Flash Lite (Reasoning)
49.0%	Adverb-first sentence starts	100Writer: Palmyra X5 99Claude Sonnet 4.5 96Claude Sonnet 4.6	0Inception Mercury 0Grok 4 Fast 0Inception Mercury 2
100.0%	Adverbs in dialogue tags	100Gemma 3 27B 100GPT-5.2 100Z.AI GLM 4.7 Flash	30Rocinante 12B 47GPT-4.1 Nano 47Hermes 3 70B
89.4%	AI-ism adverb frequency	100Qwen 3.5 122B 99Nemotron 3 Super 98Grok 4.1 Fast	70Mistral Small 3.2 24B 72Gemma 3 4B 73Llama 3.1 Nemotron 70B
100.0%	AI-ism character names	100DeepSeek V3.1 100Gemini 3.1 Flash Lite (Preview) 100Writer: Palmyra X5	92Llama 3.1 8B 96GPT-5.4 (Reasoning) 96DeepSeek V3.2
100.0%	AI-ism location names	100Z.AI GLM 5 Turbo 100Mistral Small 4 (Reasoning) 100ByteDance Seed 1.6 Flash	96Gemma 3 27B
22.8%	AI-ism word frequency	78ByteDance Seed 2.0 Lite 69GPT-5.4 66ByteDance Seed 1.6 Flash	0Grok 4 0Claude 3.5 Haiku 0GPT-4o, May 13th (temp=0)
100.0%	Cliché density	100MiniMax M2.7 100ByteDance Seed 2.0 Lite 100Claude Sonnet 4.6 (Reasoning)	7Mistral Small 3.2 24B 33Stealth: Aurora Alpha 40Inception Mercury
34.6%	Dialogue tag variety (said vs. fancy)	100Claude Sonnet 4.6 100Z.AI GLM 5 Turbo 100GPT-5.4 (Reasoning)	0Gemini 2.5 Flash Lite 0Arcee AI: Trinity Mini 0GPT-4o Mini (temp=0)
51.4%	Em-dash & semicolon overuse	100GPT-4o, May 13th (temp=0) 100Qwen 3.5 Flash 100GPT-5.4 Mini	0Mistral Medium 3.1 0Gemini 3 Flash (Preview) 0ByteDance Seed 1.6
100.0%	Emotion telling (show vs. tell)	100Stealth: Hunter Alpha 100WizardLM 2 8x22b 100Z.AI GLM 4.7	65Mistral Small 3.2 24B 65GPT-4o, Aug. 6th (temp=0) 68Llama 3.1 70B
93.7%	Filter word density	100Mistral Small Creative 100GPT-4.1 100GPT-5.4 Nano	0Inception Mercury 0Llama 3.1 Nemotron 70B 14Cohere Command R+ (Aug. 2024)
100.0%	Gibberish response detection	100Mistral Small 3.2 24B 100Grok 4.20 (Beta) 100Grok 4 Fast	40Llama 3.1 8B 80Hermes 3 70B 80Llama 3.1 70B
100.0%	Markdown formatting overuse	100GPT-4.1 100DeepSeek-V2 Chat 100Gemini 3 Flash (Preview, Reasoning)	69Ministral 3 8B 75Ministral 8B 80Mistral Medium 3.1
100.0%	Missing dialogue indicators (quotation marks)	100Aion 2.0 100Qwen 3 32B 100ByteDance Seed 2.0 Lite	40Qwen 3.5 35B 80Qwen 3.5 Flash 80ByteDance Seed 2.0 Mini
61.7%	Name drop frequency	98GPT-4.1 Nano 96Inception Mercury 96Claude Sonnet 4.6 (Reasoning)	0Qwen 3.5 9B 6Qwen 3.5 122B 14Claude 3.7 Sonnet
58.0%	Narrator intent-glossing	100o4 Mini High 100GPT-5.4 (Reasoning) 98Grok 4.1 Fast	0Claude Sonnet 4 0Claude 3.5 Haiku 0Mistral Small 3.2 24B
100.0%	Overuse of "that" (subordinate clause padding)	100Z.AI GLM 4.6 100Stealth: Healer Alpha 100Qwen 3.5 Flash	50Mistral Small 3.2 24B 80Llama 3.1 70B 80Mistral NeMO
100.0%	Paragraph length variance	100Z.AI GLM 5 Turbo 100Gemma 3 4B 100Grok 4.1 Fast	35Arcee AI: Trinity Mini 59Stealth: Aurora Alpha 65Inception Mercury 2
99.0%	Passive voice overuse	100GPT-5.4 (Reasoning) 100MiniMax M2.7 100Grok 4.20 (Beta)	84ByteDance Seed 2.0 Lite 89Llama 3.1 70B 93ByteDance Seed 2.0 Mini
100.0%	Past progressive (was/were + -ing) overuse	100Qwen3 235B A22B Instruct 2507 100Claude 3 Haiku 100GPT-4o, May 13th (temp=0)	34Ministral 8B 38Ministral 3 3B 44Gemini 3 Flash (Preview)
100.0%	Pronoun-first sentence starts	100DeepSeek V3 (2025-03-24) 100Mistral Small Creative 100Z.AI GLM 5	44ByteDance Seed 2.0 Mini 46Z.AI GLM 4.7 47Mistral NeMO
96.4%	Purple prose (modifier overload)	100Grok 4 Fast 100GPT-4o, Aug. 6th (temp=0) 100Nemotron 3 Nano	83Rocinante 12B 85Gemma 3 4B 86Gemini 3.1 Pro (Preview)
100.0%	Repeated phrase echo	100Aion 2.0 100DeepSeek V3 (2024-12-26) 100Mistral Small 4	—
100.0%	Sentence length variance	100Qwen 3.5 Plus (2026-02-15) 100Ministral 8B 100Claude 3.5 Sonnet	86GPT-4o, Aug. 6th (temp=0) 88Mistral Small 3.2 24B 90Stealth: Aurora Alpha
59.3%	Sentence opener variety	98Claude 3.5 Haiku 95Rocinante 12B 93DeepSeek V3 (2025-03-24)	31GPT-5 Nano 35Inception Mercury 37Qwen 3.5 35B
31.0%	Subject-first sentence starts	93Writer: Palmyra X5 92Rocinante 12B 85GPT-5.4 (Reasoning)	0Inception Mercury 0Qwen 3.5 122B 1Inception Mercury 2
22.1%	Subordinate conjunction sentence starts	93Gemma 3 27B 80Cohere Command R+ (Aug. 2024) 79Gemini 2.5 Flash Lite	0Inception Mercury 2 0Claude 3.5 Sonnet 0Grok 4.1 Fast
51.7%	Technical jargon density	100Gemini 3.1 Pro (Preview) 99Qwen 3.5 35B 99Hermes 3 405B	0Ministral 3B 0GPT-5 Nano 0Stealth: Aurora Alpha
40.0%	Useless dialogue additions	100Qwen 3.5 27B 100Qwen 3.5 397B A17B 100GPT-5.4 Mini (Reasoning)	0Arcee AI: Trinity Mini 0Mistral NeMO 0GPT-4o, Aug. 6th (temp=0)

Bad Writing Habits

Fantasy: entering an ancient ruin

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)