Relationship tree

Extracts a deterministic XML family and relationship tree from cumulative literary prose.

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4 (Reasoning)	96%
Claude Opus 4.6 (Reasoning)	95%
GPT-5.5 (Reasoning)	94%
Claude Opus 4.8 (Reasoning)	94%
Claude Sonnet 4.6 (Reasoning)	92%
MoonshotAI: Kimi K2.6	92%
Claude Opus 4.6	91%
Z.AI GLM 5.1	91%
Claude Opus 4.8 (Reasoning, Low)	90%
GPT-5	90%
Claude Opus 4.7 (Reasoning)	89%
Gemini 2.5 Pro	89%
Claude Opus 4.7	89%
GPT-5.4 (Reasoning, Low)	88%
MoonshotAI: Kimi K2.5	88%
GPT-5.5	88%
Gemini 3.1 Pro (Preview)	88%
MiniMax M3	87%
GPT-5.5 (Reasoning, Low)	87%
Gemini 3.5 Flash (Reasoning)	87%

Price-Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score	Cost	Time
GPT-5.5 (Reasoning, Low)	87%	$0.097	42.6s
Gemini 3 Flash (Preview)	81%	$0.0087	7.1s
Gemini 2.5 Pro	89%	$0.099	58.2s
Gemini 2.5 Flash	80%	$0.0075	6.1s
GPT-5.4 (Reasoning, Low)	88%	$0.053	35.5s
GPT-5.5	88%	$0.105	37.1s
Xiaomi MIMO v2.5	84%	$0.0027	1.1m
Gemini 3.1 Flash Lite (Preview)	78%	$0.0046	3.2s
Grok 4.20 (Beta, Reasoning)	83%	$0.032	30.1s
GPT-5.2	87%	$0.062	50.7s
Gemini 3.1 Flash Lite (Reasoning)	78%	$0.0033	3.2s
Gemma 4 26B	79%	$0.0018	27.2s
DeepSeek V4 Pro (Reasoning)	85%	$0.020	2.3m
Qwen 3.6 Flash	81%	$0.015	51.7s
Gemini 3.1 Flash Lite	76%	$0.0031	3.3s
Claude Opus 4.7 (Reasoning)	89%	$0.184	25.0s
GPT-5.4	81%	$0.033	17.4s
Gemini 3 Flash (Preview, Reasoning)	85%	$0.030	40.3s
Claude Sonnet 4.6	83%	$0.081	23.3s
DeepSeek V3.1	74%	$0.0055	52.0s

Cost vs Performance

Compares total cost for this test against the test score. Quadrant lines are drawn at the median values. Only models with available cost data are shown.

2 low-scoring outliers hidden: Llama 3.1 8B (38.0%), Gemma 3 12B (37.1%).

Most Stable Models (Top 20)

Ranked by stability (median × consistency). Click a model name to view its detail page.

	Score	Consistency	Stability
GPT-5.4 (Reasoning)	96%	93%	90%
Claude Opus 4.6 (Reasoning)	95%	93%	88%
GPT-5.5 (Reasoning)	94%	93%	87%
Claude Opus 4.8 (Reasoning)	94%	90%	86%
Claude Sonnet 4.6 (Reasoning)	92%	88%	83%
MoonshotAI: Kimi K2.6	92%	89%	82%
GPT-5.4 (Reasoning, Low)	88%	93%	80%
Claude Opus 4.6	91%	90%	80%
GPT-5.2	87%	92%	79%
Z.AI GLM 5.1	91%	88%	79%
MiniMax M3	87%	91%	78%
Claude Opus 4.5	85%	92%	78%
Claude Opus 4.8 (Reasoning, Low)	90%	86%	77%
Z.AI GLM 5 Turbo	87%	86%	77%
GPT-5.4	81%	95%	77%
Grok 4.20 (Beta, Reasoning)	83%	92%	77%
Claude Sonnet 4.5	79%	97%	77%
Claude Opus 4.7 (Reasoning)	89%	83%	76%
Qwen 3.6 Flash	81%	94%	76%
Gemini 3 Flash (Preview)	81%	94%	76%

Top Overall Models (Top 20)

Ranked by composite score (performance, cost, speed & stability). Click a model name to view its detail page.

	Score	Cost	Speed	Stability
GPT-5.4 (Reasoning, Low)	88%	$0.053	35.5s	80%
Gemini 3 Flash (Preview)	81%	$0.0087	7.1s	76%
Gemini 3.1 Flash Lite (Preview)	78%	$0.0046	3.2s	76%
Gemini 2.5 Flash	80%	$0.0075	6.1s	75%
GPT-5.2	87%	$0.062	50.7s	79%
Grok 4.20 (Beta, Reasoning)	83%	$0.032	30.1s	77%
Gemma 4 26B	79%	$0.0018	27.2s	76%
GPT-5.4	81%	$0.033	17.4s	77%
Gemini 3.1 Flash Lite (Reasoning)	78%	$0.0033	3.2s	74%
Qwen 3.6 Flash	81%	$0.015	51.7s	76%
Xiaomi MIMO v2.5	84%	$0.0027	1.1m	73%
Claude Opus 4.6	91%	$0.154	31.4s	80%
Gemini 3.1 Flash Lite	76%	$0.0031	3.3s	71%
GPT-5.4 (Reasoning)	96%	$0.175	2.6m	90%
Claude Sonnet 4.6	83%	$0.081	23.3s	76%
GPT-5.5	88%	$0.105	37.1s	75%
Inception Mercury 2	76%	$0.0096	12.8s	71%
ByteDance Seed 1.6	81%	$0.014	1.2m	72%
Grok 4.20 (Reasoning)	81%	$0.027	40.0s	71%
GPT-5 Mini	84%	$0.019	2.0m	75%

Model	Total ▼	Core relationship tree	Family relationship tree
GPT-5.4 (Reasoning)	96%	99%	93%
Claude Opus 4.6 (Reasoning)	95%	98%	92%
GPT-5.5 (Reasoning)	94%	98%	90%
Claude Opus 4.8 (Reasoning)	94%	99%	89%
Claude Sonnet 4.6 (Reasoning)	92%	98%	87%
MoonshotAI: Kimi K2.6	92%	97%	88%
Claude Opus 4.6	91%	95%	88%
Z.AI GLM 5.1	91%	96%	87%
Claude Opus 4.8 (Reasoning, Low)	90%	90%	90%
GPT-5	90%	98%	82%
Claude Opus 4.7 (Reasoning)	89%	97%	82%
Gemini 2.5 Pro	89%	97%	81%
Claude Opus 4.7	89%	95%	83%
GPT-5.4 (Reasoning, Low)	88%	91%	85%
MoonshotAI: Kimi K2.5	88%	93%	83%

1–15 of 140

Page 1 / 10

Rows:

Core relationship tree

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4 (Reasoning)	99%
Claude Opus 4.8 (Reasoning)	99%
Claude Sonnet 4.6 (Reasoning)	98%
Claude Opus 4.6 (Reasoning)	98%
GPT-5.5 (Reasoning)	98%
GPT-5	98%
Gemini 2.5 Pro	97%
GPT-5.5 (Reasoning, Low)	97%
MoonshotAI: Kimi K2.6	97%
Gemini 3.1 Pro (Preview)	97%
Claude Opus 4.7 (Reasoning)	97%
Z.AI GLM 5.1	96%
Claude Opus 4.7	95%
Claude Opus 4.6	95%
GPT-5.5	94%
MoonshotAI: Kimi K2.5	93%
Qwen3.7 Max	92%
Z.AI GLM 5 Turbo	92%
GPT-5.4 Mini (Reasoning)	92%
Qwen3.6 Max Preview	91%

Price-Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score	Cost	Time
Gemini 3 Flash (Preview, Reasoning)	90%	$0.024	27.7s
Xiaomi MIMO v2.5	88%	$0.0017	47.0s
Gemini 3 Flash (Preview)	83%	$0.0065	5.0s
GPT-5.4 (Reasoning, Low)	91%	$0.033	20.9s
GPT-5 Mini	87%	$0.013	1.4m
Gemma 4 26B	81%	$0.0010	22.3s
DeepSeek V4 Flash (Reasoning)	84%	$0.0022	1.9m
Z.AI GLM 5 Turbo	92%	$0.019	1.7m
Gemini 3.5 Flash (Reasoning, Minimal)	83%	$0.021	5.7s
ByteDance Seed 1.6	85%	$0.010	59.2s
Gemini 2.5 Flash	81%	$0.0050	5.3s
Gemini 3.1 Flash Lite (Preview)	79%	$0.0029	2.6s
GPT-5.2	90%	$0.039	32.6s
GPT-5.4 Mini (Reasoning, Low)	76%	$0.0076	9.4s
GPT-4o, Aug. 6th (temp=0)	81%	$0.020	7.2s
DeepSeek V3.1	73%	$0.0028	37.1s
DeepSeek V4 Pro (Reasoning)	88%	$0.011	1.4m
GPT-5.1	82%	$0.016	11.2s
Grok 4.20 (Beta, Reasoning)	86%	$0.023	30.1s
Gemini 3.1 Flash Lite (Reasoning)	79%	$0.0024	2.4s

Most Stable Models (Top 20)

Ranked by stability (median × consistency). Click a model name to view its detail page.

	Score	Consistency	Stability
Claude Sonnet 4.6 (Reasoning)	98%	100%	98%
Claude Opus 4.8 (Reasoning)	99%	99%	97%
GPT-5.5 (Reasoning)	98%	98%	96%
GPT-5.5 (Reasoning, Low)	97%	98%	96%
Claude Opus 4.6 (Reasoning)	98%	97%	96%
GPT-5.4 (Reasoning)	99%	98%	96%
Gemini 3.1 Pro (Preview)	97%	98%	95%
GPT-5	98%	97%	94%
MoonshotAI: Kimi K2.6	97%	95%	93%
Claude Opus 4.7 (Reasoning)	97%	97%	93%
Gemini 2.5 Pro	97%	95%	93%
GPT-5.5	94%	96%	91%
Claude Opus 4.7	95%	92%	90%
Z.AI GLM 5 Turbo	92%	96%	89%
Z.AI GLM 5.1	96%	89%	88%
Claude Opus 4.6	95%	90%	86%
GPT-5.2	90%	95%	85%
Qwen3.7 Max	92%	88%	85%
GPT-5.4 (Reasoning, Low)	91%	94%	85%
MiniMax M3	90%	91%	84%

Top Overall Models (Top 20)

Ranked by composite score (performance, cost, speed & stability). Click a model name to view its detail page.

	Score	Cost	Speed	Stability
GPT-5.5 (Reasoning, Low)	97%	$0.070	24.6s	96%
GPT-5.5	94%	$0.062	16.3s	91%
GPT-5.4 (Reasoning, Low)	91%	$0.033	20.9s	85%
Gemini 3 Flash (Preview)	83%	$0.0065	5.0s	81%
Z.AI GLM 5 Turbo	92%	$0.019	1.7m	89%
Xiaomi MIMO v2.5	88%	$0.0017	47.0s	78%
GPT-5.2	90%	$0.039	32.6s	85%
Gemini 2.5 Pro	97%	$0.082	51.0s	93%
Grok 4.20 (Beta, Reasoning)	86%	$0.023	30.1s	83%
Gemma 4 26B	81%	$0.0010	22.3s	80%
Gemini 3 Flash (Preview, Reasoning)	90%	$0.024	27.7s	77%
Gemini 2.5 Flash	81%	$0.0050	5.3s	78%
GPT-5.4 (Reasoning)	99%	$0.092	1.4m	96%
Claude Opus 4.7 (Reasoning)	97%	$0.111	17.6s	93%
ByteDance Seed 1.6	85%	$0.010	59.2s	80%
Gemini 3.1 Flash Lite (Preview)	79%	$0.0029	2.6s	76%
GPT-5.4	82%	$0.020	12.5s	79%
Claude Opus 4.6	95%	$0.088	19.0s	86%
GPT-5	98%	$0.077	2.0m	94%
Grok 4.20 (Reasoning)	85%	$0.024	43.9s	80%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
13	GPT-5.4 (Reasoning)	$0.092	1.4m	96%	100	100	98	98	98	99%
53	Claude Opus 4.8 (Reasoning)	$0.177	43.1s	97%	100	98	98	98	98	99%
105	Claude Sonnet 4.6 (Reasoning)	$0.209	2.7m	98%	98	98	98	98	98	98%
73	Claude Opus 4.6 (Reasoning)	$0.181	1.2m	96%	100	99	98	98	96	98%
38	GPT-5.5 (Reasoning)	$0.132	1.3m	96%	98	98	98	96	96	98%
19	GPT-5	$0.077	2.0m	94%	100	98	97	96	96	98%
8	Gemini 2.5 Pro	$0.082	51.0s	93%	100	98	98	98	92	97%
1	GPT-5.5 (Reasoning, Low)	$0.070	24.6s	96%	98	98	98	96	96	97%
22	MoonshotAI: Kimi K2.6	$0.055	2.7m	93%	100	98	98	96	93	97%
40	Gemini 3.1 Pro (Preview)	$0.130	1.2m	95%	98	98	96	96	96	97%
14	Claude Opus 4.7 (Reasoning)	$0.111	17.6s	93%	98	98	96	96	96	97%
92	Z.AI GLM 5.1	$0.060	5.3m	88%	98	98	98	98	85	96%
25	Claude Opus 4.7	$0.112	17.7s	90%	98	98	98	93	89	95%
18	Claude Opus 4.6	$0.088	19.0s	86%	99	98	96	96	85	95%
2	GPT-5.5	$0.062	16.3s	91%	97	95	95	93	92	94%
46	MoonshotAI: Kimi K2.5	$0.023	3.5m	82%	100	98	94	86	85	93%
32	Qwen3.7 Max	$0.034	2.5m	85%	98	96	96	85	85	92%
5	Z.AI GLM 5 Turbo	$0.019	1.7m	89%	95	93	93	91	89	92%
43	GPT-5.4 Mini (Reasoning)	$0.067	1.6m	83%	99	93	91	89	87	92%
84	Qwen3.6 Max Preview	$0.070	3.5m	83%	98	90	90	90	87	91%
3	GPT-5.4 (Reasoning, Low)	$0.033	20.9s	85%	95	93	90	90	86	91%
11	Gemini 3 Flash (Preview, Reasoning)	$0.024	27.7s	77%	100	98	97	83	73	90%
62	MiniMax M3	$0.016	4.3m	84%	95	93	93	85	84	90%
116	Claude Opus 4.8 (Reasoning, Low)	$0.192	51.0s	76%	98	98	93	86	73	90%
7	GPT-5.2	$0.039	32.6s	85%	93	92	90	87	87	90%
63	Gemini 3.5 Flash (Reasoning)	$0.090	40.4s	75%	98	98	97	86	69	90%
74	Z.AI GLM 5	$0.027	3.2m	75%	98	98	90	80	79	89%
23	o4 Mini High	$0.045	1.0m	84%	91	90	88	87	85	88%
29	DeepSeek V4 Pro (Reasoning)	$0.011	1.4m	74%	100	90	85	83	82	88%
6	Xiaomi MIMO v2.5	$0.0017	47.0s	78%	97	89	87	84	83	88%
69	Qwen 3.5 Plus (2026-02-15)	$0.018	3.7m	81%	93	90	88	84	81	87%
45	Claude Opus 4.5	$0.077	13.4s	77%	96	87	84	84	84	87%
27	Claude Sonnet 4.6	$0.049	16.0s	79%	94	87	85	85	84	87%
26	GPT-5 Mini	$0.013	1.4m	79%	92	90	87	85	78	87%
9	Grok 4.20 (Beta, Reasoning)	$0.023	30.1s	83%	89	87	86	85	84	86%
20	Grok 4.20 (Reasoning)	$0.024	43.9s	80%	90	87	85	82	82	85%
15	ByteDance Seed 1.6	$0.010	59.2s	80%	90	87	86	83	80	85%
70	Qwen 3.6 27B	$0.030	2.6m	77%	91	86	83	83	82	85%
91	ByteDance Seed 2.0 Lite	$0.011	3.4m	69%	97	90	82	79	75	85%
80	Qwen 3.5 397B A17B	$0.031	2.9m	75%	90	88	87	86	71	85%
54	DeepSeek V4 Flash (Reasoning)	$0.0022	1.9m	69%	98	87	82	78	76	84%
58	Qwen 3.5 Flash	$0.0032	3.6m	82%	85	85	85	82	81	84%
57	Qwen 3.5 Plus (2026-04-20)	$0.018	2.5m	79%	88	84	83	83	80	84%
4	Gemini 3 Flash (Preview)	$0.0065	5.0s	81%	86	85	84	82	80	83%
31	Gemini 3.5 Flash (Reasoning, Minimal)	$0.021	5.7s	70%	96	89	84	76	72	83%
98	Gemma 4 31B (Reasoning)	$0.0031	4.2m	70%	90	90	87	85	64	83%
56	Qwen 3.5 122B	$0.031	2.3m	81%	85	84	83	83	81	83%
86	Z.AI GLM 4.7	$0.017	3.2m	74%	90	89	87	78	71	83%
28	Qwen 3.5 35B	$0.014	1.4m	81%	84	84	84	82	80	83%
24	Qwen 3.6 Flash	$0.012	40.4s	77%	86	85	81	81	79	82%
17	GPT-5.4	$0.020	12.5s	79%	84	84	82	81	78	82%
35	GPT-5.1	$0.016	11.2s	69%	92	86	81	80	70	82%
39	o4 Mini	$0.022	29.5s	73%	87	85	85	84	67	82%
47	Xiaomi MIMO v2.5 Pro	$0.0052	1.2m	68%	92	87	80	79	70	81%
33	GPT-4o, Aug. 6th (temp=0)	$0.020	7.2s	72%	89	87	84	77	69	81%
12	Gemini 2.5 Flash	$0.0050	5.3s	78%	84	81	80	80	79	81%
30	Qwen 3.6 35B	$0.012	59.4s	79%	82	82	82	81	77	81%
68	Z.AI GLM 4.6	$0.011	2.4m	75%	85	84	81	80	74	81%
10	Gemma 4 26B	$0.0010	22.3s	80%	81	81	81	80	80	81%
42	GPT-OSS 120B	$0.0014	1.4m	74%	87	83	83	78	71	80%
123	ByteDance Seed 2.0 Mini	$0.0047	7.0m	72%	86	84	82	73	71	79%
72	Gemma 4 26B (Reasoning)	$0.0038	2.5m	74%	84	81	79	77	73	79%
16	Gemini 3.1 Flash Lite (Preview)	$0.0029	2.6s	76%	81	81	79	78	76	79%
55	GPT-4o, May 13th (temp=0)	$0.054	4.2s	75%	82	82	82	76	71	79%
21	Gemini 3.1 Flash Lite (Reasoning)	$0.0024	2.4s	74%	82	80	77	77	77	79%
110	Qwen 3.5 27B	$0.018	2.7m	57%	91	85	85	85	46	78%
49	Claude Sonnet 4.5	$0.045	12.2s	75%	81	80	77	77	77	78%
139	Claude Opus 4	$0.222	4.2m	68%	84	83	82	80	62	78%
51	Gemma 4 31B	$0.0015	1.5m	73%	84	80	79	74	73	78%
36	Inception Mercury 2	$0.0066	9.0s	71%	84	80	76	75	75	78%
37	DeepSeek V4 Pro	$0.0048	12.6s	70%	85	80	77	75	73	78%
52	DeepSeek V4 Flash	$0.0012	1.5m	72%	83	80	78	75	72	78%
61	GPT-4o, May 13th (temp=1)	$0.050	4.0s	73%	82	79	79	76	70	77%
34	Gemini 3.1 Flash Lite	$0.0026	2.4s	71%	82	78	77	76	69	76%
44	GPT-5.4 Nano (Reasoning)	$0.0050	24.4s	68%	85	79	77	71	69	76%
50	GPT-5.4 Mini (Reasoning, Low)	$0.0076	9.4s	65%	83	81	81	79	56	76%
48	DeepSeek V3.2	$0.0023	26.8s	67%	86	76	75	73	69	76%
41	GPT-4.1	$0.015	6.2s	73%	79	78	78	73	71	76%
59	Z.AI GLM 4.5	$0.0062	49.2s	67%	82	81	78	73	64	76%
83	Claude Sonnet 4	$0.044	12.6s	66%	78	77	77	77	60	74%
79	DeepSeek V3.1	$0.0028	37.1s	58%	81	81	80	79	46	73%
66	DeepSeek-V2 Chat	$0.0024	27.3s	64%	79	78	71	69	67	73%
60	Grok 4.3	$0.0071	7.2s	65%	80	76	75	71	60	72%
65	GPT-4o, Aug. 6th (temp=1)	$0.019	7.2s	68%	76	73	73	71	66	72%
75	Mistral Large 2	$0.022	13.6s	68%	75	71	71	70	69	71%
71	Grok 4.20	$0.010	5.3s	64%	75	75	69	67	66	71%
82	Grok 4.20 (Beta)	$0.0100	4.3s	59%	83	71	67	66	65	70%
67	DeepSeek V3 (2024-12-26)	$0.0025	22.7s	66%	74	70	69	69	69	70%
85	MiniMax M2.5	$0.0032	37.0s	61%	77	76	72	70	55	70%
96	Grok 4.3 (Reasoning)	$0.0073	13.7s	54%	83	78	72	70	46	70%
89	Mistral Large	$0.022	40.1s	67%	72	72	71	70	65	70%
78	Mistral Medium 3.1	$0.0049	9.5s	60%	78	73	71	71	54	69%
76	Claude Haiku 4.5	$0.014	5.7s	64%	75	74	72	63	63	69%
64	Mistral Large 3	$0.0055	12.4s	68%	71	71	71	68	65	69%
138	Qwen 3.5 9B	$0.0035	7.9m	59%	76	74	67	67	61	69%
102	GPT-5 Nano	$0.0067	1.9m	63%	74	74	72	64	57	68%
95	Mistral Small 4 (Reasoning)	$0.0035	20.7s	57%	78	75	74	64	46	67%
94	Hermes 3 405B	$0.0085	43.7s	63%	71	71	69	66	60	67%
109	Gemini 2.5 Flash (Reasoning)	$0.018	28.2s	51%	86	64	63	62	60	67%
108	WizardLM 2 8x22b	$0.0055	1.6m	59%	75	73	72	60	52	66%
77	Gemini 2.5 Flash Lite	$0.0012	2.9s	63%	69	68	66	65	63	66%
88	Ministral 3 8B	$0.0014	11.0s	61%	71	68	66	63	61	66%
101	Gemini 2.5 Flash Lite (Reasoning)	$0.0055	41.0s	55%	74	74	67	61	51	65%
119	Z.AI GLM 4.7 Flash	$0.0044	2.7m	52%	78	68	61	61	59	65%
87	Ministral 8B	$0.0009	9.2s	61%	69	67	65	63	62	65%
90	Llama 3.1 70B	$0.0042	22.3s	63%	68	67	67	63	60	65%
99	Qwen 2.5 72B	$0.0031	43.5s	57%	72	70	65	64	54	65%
81	Mistral Small 3.2 24B	$0.0012	10.2s	63%	67	65	65	65	62	65%
93	GPT-5.4 Mini	$0.0054	6.2s	60%	70	67	64	62	61	65%
106	MiniMax M2.7	$0.0044	14.7s	49%	76	75	60	59	52	64%
104	Qwen3 235B A22B Instruct 2507	$0.0009	44.0s	55%	74	67	62	62	56	64%
97	Writer: Palmyra X5	$0.010	13.3s	60%	68	67	64	60	59	64%
117	Z.AI GLM 4.5 Air	$0.0043	1.7m	51%	74	74	70	59	39	63%
100	GPT-4.1 Mini	$0.0032	22.0s	54%	74	66	62	57	57	63%
136	Aion 2.0	$0.016	2.4m	29%	86	83	46	46	46	62%
111	Claude 3 Haiku	$0.0034	8.6s	50%	76	60	60	57	53	61%
107	GPT-5.4 Nano (Reasoning, Low)	$0.0025	10.7s	52%	70	67	61	55	51	61%
130	Nemotron 3 Nano	$0.0033	2.9m	48%	71	70	62	57	39	60%
113	Gemma 3 27B	$0.0007	21.3s	52%	69	66	61	56	46	60%
114	DeepSeek V3 (2025-03-24)	$0.0025	32.0s	51%	69	69	65	46	46	59%
103	Ministral 3 14B	$0.0015	11.2s	56%	62	59	58	57	57	59%
112	GPT-4.1 Nano	$0.0005	5.5s	51%	66	64	59	58	47	59%
126	Qwen 3 32B	$0.0032	2.4m	50%	71	60	60	53	48	59%
118	Mistral Small 4	$0.0017	5.7s	42%	71	71	53	51	45	58%
125	ByteDance Seed 1.6 Flash	$0.0019	38.2s	43%	67	63	53	45	40	54%
115	GPT-5.4 Nano	$0.0015	8.5s	52%	56	55	54	53	49	53%
121	Hermes 3 70B	$0.0026	29.1s	46%	64	58	55	46	41	53%
120	Mistral NeMO	$0.0014	11.8s	47%	53	52	49	48	47	50%
122	Arcee AI: Trinity Mini	$0.0005	7.0s	46%	54	53	50	45	45	49%
129	Cydonia 24B V4.1	$0.0018	24.8s	39%	63	56	48	42	36	49%
132	Cohere Command R+ (Aug. 2024)	$0.028	22.0s	44%	55	50	49	47	38	48%
127	Ministral 3 3B	$0.0003	5.3s	42%	54	54	48	41	39	47%
128	Gemma 3 4B	$0.0004	20.5s	44%	51	50	46	45	43	47%
124	LFM2 24B	$0.0004	7.8s	46%	46	46	46	46	46	46%
131	Rocinante 12B	$0.0020	28.3s	41%	51	46	45	43	36	44%
133	Ministral 3B	$0.0004	6.3s	36%	52	51	41	41	36	44%
135	Llama 3.1 8B	$0.0006	25.6s	30%	64	41	39	38	30	43%
134	Skyfall 36B V2	$0.0036	24.5s	38%	50	46	42	39	35	42%
137	Gemma 3 12B	$0.0003	24.3s	32%	46	37	35	35	33	37%
75.28%

Median	Evaluator	Top 3	Flop 3
78.2%	Alias accuracy	100Qwen 3.5 Flash 100GPT-5.5 (Reasoning) 100GPT-5 Mini	0LFM2 24B 2Arcee AI: Trinity Mini 2Gemma 3 12B
100.0%	Character precision	100Grok 4.3 100Claude Sonnet 4.6 (Reasoning) 100Gemma 4 31B	20Gemma 3 12B 56Llama 3.1 8B 61Gemini 2.5 Flash (Reasoning)
100.0%	Character recall	100Qwen 3.5 122B 100GPT-5.2 100Xiaomi MIMO v2.5	0LFM2 24B 0Gemma 3 12B 40Aion 2.0
100.0%	Isolated character handling	100Gemini 2.5 Flash (Reasoning) 100MiniMax M3 100GPT-5.4 Mini (Reasoning, Low)	0LFM2 24B 1Gemma 3 12B 3Skyfall 36B V2
100.0%	Red-herring resistance	100GPT-5.5 (Reasoning, Low) 100Z.AI GLM 4.7 Flash 100Claude Sonnet 4.5	63Ministral 3B 81Mistral Small 4 81Rocinante 12B
21.0%	Relationship category recall	100Claude Opus 4.6 (Reasoning) 100GPT-5.5 100Claude Opus 4.7 (Reasoning)	0LFM2 24B 0Gemma 3 12B 0Gemma 3 4B
100.0%	Relationship endpoint integrity	100Grok 4.20 100DeepSeek-V2 Chat 100GPT-5	16Mistral Small 4 20Gemma 3 12B 40Cohere Command R+ (Aug. 2024)
81.1%	Relationship precision	100LFM2 24B 100Claude Sonnet 4.5 100Z.AI GLM 5	4Gemma 3 4B 5Mistral NeMO 10Ministral 3B
17.2%	Relationship recall	100Gemini 3.1 Pro (Preview) 100GPT-5.5 100GPT-5.4 (Reasoning)	0LFM2 24B 0Gemma 3 12B 0Gemma 3 4B
78.6%	Relationship type accuracy	100Gemma 3 12B 100Mistral Large 2 100GPT-5.4 (Reasoning)	36Rocinante 12B 36Ministral 3B 36Hermes 3 70B
100.0%	XML structure	100ByteDance Seed 2.0 Mini 100Gemini 2.5 Flash Lite 100Gemma 4 26B (Reasoning)	10LFM2 24B 46Aion 2.0 55Rocinante 12B

Family relationship tree

Tooling Reasoning Hallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4 (Reasoning)	93%
Claude Opus 4.6 (Reasoning)	92%
GPT-5.5 (Reasoning)	90%
Claude Opus 4.8 (Reasoning, Low)	90%
Claude Opus 4.8 (Reasoning)	89%
Claude Opus 4.6	88%
MoonshotAI: Kimi K2.6	88%
Z.AI GLM 5.1	87%
Claude Sonnet 4.6 (Reasoning)	87%
GPT-5.4 (Reasoning, Low)	85%
MiniMax M3	85%
DeepSeek V4 Flash (Reasoning)	84%
Gemini 3.5 Flash (Reasoning)	84%
GPT-5.2	84%
Claude Opus 4.5	83%
MoonshotAI: Kimi K2.5	83%
Claude Opus 4.7	83%
GPT-5	82%
DeepSeek V4 Pro (Reasoning)	82%
GPT-5.4 Mini (Reasoning)	82%

Price-Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score	Cost	Time
Gemini 2.5 Flash	78%	$0.010	6.9s
Gemini 3.1 Flash Lite (Preview)	78%	$0.0062	3.9s
GPT-5.4 (Reasoning, Low)	85%	$0.072	50.1s
Gemini 3 Flash (Preview)	79%	$0.011	9.2s
Gemini 3.1 Flash Lite	76%	$0.0037	4.2s
Qwen 3.6 Flash	80%	$0.019	1.1m
Grok 4.20 (Beta, Reasoning)	80%	$0.041	30.2s
Gemini 3.1 Flash Lite (Reasoning)	77%	$0.0042	3.9s
GPT-5.4	80%	$0.047	22.4s
Gemini 3 Flash (Preview, Reasoning)	79%	$0.037	53.0s
Grok 4.3	75%	$0.012	8.3s
GPT-5.4 Mini (Reasoning, Low)	75%	$0.020	19.5s
DeepSeek V3.2	78%	$0.0070	40.9s
Gemma 4 26B	77%	$0.0026	32.0s
GPT-5.4 Nano (Reasoning)	78%	$0.013	59.0s
DeepSeek V4 Flash	74%	$0.0028	14.4s
Grok 4.3 (Reasoning)	75%	$0.018	14.1s
GPT-5.2	84%	$0.086	1.1m
Inception Mercury 2	75%	$0.013	16.6s
Gemma 4 31B	79%	$0.0040	2.3m

Most Stable Models (Top 20)

Ranked by stability (median × consistency). Click a model name to view its detail page.

	Score	Consistency	Stability
Claude Opus 4.6 (Reasoning)	92%	98%	90%
GPT-5.4 (Reasoning)	93%	95%	88%
GPT-5.5 (Reasoning)	90%	98%	88%
Claude Opus 4.6	88%	97%	85%
Claude Opus 4.8 (Reasoning, Low)	90%	95%	85%
GPT-5.4 (Reasoning, Low)	85%	98%	83%
MoonshotAI: Kimi K2.6	88%	94%	83%
Claude Opus 4.8 (Reasoning)	89%	95%	83%
GPT-5.2	84%	97%	82%
Z.AI GLM 5.1	87%	94%	82%
DeepSeek V4 Flash (Reasoning)	84%	97%	82%
Gemini 3.5 Flash (Reasoning)	84%	97%	82%
Claude Sonnet 4.6 (Reasoning)	87%	95%	81%
GPT-5	82%	99%	81%
MiniMax M3	85%	94%	80%
Claude Opus 4.5	83%	95%	80%
Qwen3.6 Max Preview	82%	97%	80%
MoonshotAI: Kimi K2.5	83%	92%	78%
Gemma 4 31B	79%	99%	78%
GPT-5.5	81%	96%	78%

Top Overall Models (Top 20)

Ranked by composite score (performance, cost, speed & stability). Click a model name to view its detail page.

	Score	Cost	Speed	Stability
GPT-5.4 (Reasoning, Low)	85%	$0.072	50.1s	83%
Gemini 3 Flash (Preview)	79%	$0.011	9.2s	76%
Gemini 3.1 Flash Lite (Preview)	78%	$0.0062	3.9s	76%
GPT-5.2	84%	$0.086	1.1m	82%
Gemini 2.5 Flash	78%	$0.010	6.9s	75%
Qwen 3.6 Flash	80%	$0.019	1.1m	77%
DeepSeek V3.2	78%	$0.0070	40.9s	76%
Grok 4.20 (Beta, Reasoning)	80%	$0.041	30.2s	76%
Claude Opus 4.6	88%	$0.221	43.8s	85%
GPT-5.4	80%	$0.047	22.4s	76%
Gemini 3.5 Flash (Reasoning)	84%	$0.136	58.4s	82%
Gemini 3.1 Flash Lite (Reasoning)	77%	$0.0042	3.9s	73%
Gemma 4 26B	77%	$0.0026	32.0s	74%
GPT-5.4 Nano (Reasoning)	78%	$0.013	59.0s	74%
Gemini 3.1 Flash Lite	76%	$0.0037	4.2s	72%
Gemma 4 31B	79%	$0.0040	2.3m	78%
Gemini 3 Flash (Preview, Reasoning)	79%	$0.037	53.0s	75%
Grok 4.20 (Reasoning)	76%	$0.030	36.0s	75%
Inception Mercury 2	75%	$0.013	16.6s	73%
GPT-5 Mini	81%	$0.024	2.5m	78%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
36	GPT-5.4 (Reasoning)	$0.259	3.7m	88%	96	95	93	92	89	93%
90	Claude Opus 4.6 (Reasoning)	$0.530	4.0m	90%	93	92	92	91	91	92%
57	GPT-5.5 (Reasoning)	$0.369	2.7m	88%	93	90	90	90	89	90%
99	Claude Opus 4.8 (Reasoning, Low)	$0.573	2.9m	85%	95	90	90	90	87	90%
104	Claude Opus 4.8 (Reasoning)	$0.585	3.1m	83%	94	89	88	87	86	89%
9	Claude Opus 4.6	$0.221	43.8s	85%	90	88	88	88	85	88%
64	MoonshotAI: Kimi K2.6	$0.109	7.2m	83%	92	89	88	86	84	88%
84	Z.AI GLM 5.1	$0.102	9.3m	82%	91	88	87	86	83	87%
135	Claude Sonnet 4.6 (Reasoning)	$0.688	9.1m	81%	91	87	86	86	83	87%
1	GPT-5.4 (Reasoning, Low)	$0.072	50.1s	83%	87	85	85	85	85	85%
115	MiniMax M3	$0.055	14.3m	80%	88	87	85	82	81	85%
28	DeepSeek V4 Flash (Reasoning)	$0.0060	5.1m	82%	87	84	84	84	83	84%
11	Gemini 3.5 Flash (Reasoning)	$0.136	58.4s	82%	86	85	84	84	82	84%
4	GPT-5.2	$0.086	1.1m	82%	85	85	84	83	81	84%
24	Claude Opus 4.5	$0.185	26.2s	80%	87	84	84	82	80	83%
69	MoonshotAI: Kimi K2.5	$0.039	7.3m	78%	88	85	85	80	77	83%
50	Claude Opus 4.7	$0.252	30.9s	78%	86	85	83	82	77	83%
33	GPT-5	$0.107	2.6m	81%	84	83	82	82	82	82%
37	DeepSeek V4 Pro (Reasoning)	$0.030	3.3m	75%	89	82	81	81	78	82%
73	GPT-5.4 Mini (Reasoning)	$0.168	4.0m	75%	88	87	86	79	70	82%
47	Qwen3.6 Max Preview	$0.098	3.9m	80%	84	83	82	81	80	82%
61	Claude Opus 4.7 (Reasoning)	$0.256	32.4s	74%	89	87	84	76	75	82%
75	Z.AI GLM 5 Turbo	$0.068	5.4m	72%	90	86	83	77	72	82%
38	Aion 2.0	$0.031	3.6m	78%	84	83	83	79	76	81%
35	GPT-5.5	$0.148	57.9s	78%	83	83	82	81	77	81%
20	GPT-5 Mini	$0.024	2.5m	78%	84	83	82	78	77	81%
42	Gemini 2.5 Pro	$0.116	1.1m	72%	87	86	80	78	72	81%
97	ByteDance Seed 2.0 Mini	$0.0077	10.4m	77%	83	82	80	80	78	81%
6	Qwen 3.6 Flash	$0.019	1.1m	77%	82	82	81	79	76	80%
56	Qwen 3.5 27B	$0.040	4.7m	78%	81	81	80	79	77	80%
25	Xiaomi MIMO v2.5	$0.0037	1.5m	71%	86	82	78	78	74	80%
10	GPT-5.4	$0.047	22.4s	76%	83	81	79	79	76	80%
8	Grok 4.20 (Beta, Reasoning)	$0.041	30.2s	76%	82	82	81	78	75	80%
44	Qwen 3.5 Plus (2026-02-15)	$0.023	4.2m	78%	81	81	80	79	77	80%
101	Claude Opus 4	$0.514	36.8s	77%	81	80	79	79	78	79%
30	Claude Sonnet 4.6	$0.114	30.6s	76%	83	80	79	78	77	79%
21	Claude Sonnet 4.5	$0.110	25.0s	77%	81	79	79	78	78	79%
49	Xiaomi MIMO v2.5 Pro	$0.010	2.7m	69%	88	83	80	78	67	79%
2	Gemini 3 Flash (Preview)	$0.011	9.2s	76%	82	79	79	78	77	79%
52	ByteDance Seed 2.0 Lite	$0.016	3.8m	74%	81	81	81	80	71	79%
17	Gemini 3 Flash (Preview, Reasoning)	$0.037	53.0s	75%	83	81	80	76	74	79%
16	Gemma 4 31B	$0.0040	2.3m	78%	79	79	79	79	78	79%
14	GPT-5.4 Nano (Reasoning)	$0.013	59.0s	74%	82	79	79	78	73	78%
7	DeepSeek V3.2	$0.0070	40.9s	76%	81	79	78	77	76	78%
45	Qwen 3.5 397B A17B	$0.058	2.3m	73%	82	80	79	79	71	78%
95	Gemini 3.1 Pro (Preview)	$0.214	1.7m	62%	88	86	73	72	72	78%
26	Claude Sonnet 4	$0.100	22.5s	77%	79	78	78	77	77	78%
5	Gemini 2.5 Flash	$0.010	6.9s	75%	81	81	81	76	71	78%
3	Gemini 3.1 Flash Lite (Preview)	$0.0062	3.9s	76%	79	79	79	77	75	78%
60	Z.AI GLM 4.7	$0.033	4.1m	74%	81	80	77	76	75	78%
54	Qwen 3.5 Plus (2026-04-20)	$0.028	3.6m	75%	80	79	79	77	72	78%
83	GPT-5.5 (Reasoning, Low)	$0.124	1.0m	57%	92	84	83	83	46	78%
65	Qwen 3.6 27B	$0.038	2.7m	67%	83	83	82	79	59	77%
31	ByteDance Seed 1.6	$0.018	1.4m	73%	81	78	78	77	72	77%
12	Gemini 3.1 Flash Lite (Reasoning)	$0.0042	3.9s	73%	79	79	76	75	75	77%
13	Gemma 4 26B	$0.0026	32.0s	74%	79	77	77	76	75	77%
51	o4 Mini High	$0.072	1.5m	72%	82	77	77	76	71	77%
82	Qwen 3.5 122B	$0.039	5.8m	72%	79	79	79	77	68	77%
18	Grok 4.20 (Reasoning)	$0.030	36.0s	75%	77	77	76	76	75	76%
15	Gemini 3.1 Flash Lite	$0.0037	4.2s	72%	80	78	76	76	71	76%
70	Z.AI GLM 4.6	$0.025	3.7m	68%	82	80	75	72	71	76%
48	o4 Mini	$0.053	1.0m	69%	83	76	76	73	71	76%
106	Z.AI GLM 5	$0.047	5.9m	58%	85	84	81	81	46	76%
23	Grok 4.3 (Reasoning)	$0.018	14.1s	72%	79	78	77	73	70	75%
41	Mistral Large 2	$0.051	27.4s	70%	80	77	75	74	70	75%
27	GPT-5.4 Mini (Reasoning, Low)	$0.020	19.5s	72%	79	77	77	72	70	75%
19	Inception Mercury 2	$0.013	16.6s	73%	76	76	75	75	72	75%
55	GPT-OSS 120B	$0.0065	3.1m	73%	77	75	74	74	74	75%
71	Qwen 3.5 Flash	$0.0047	4.0m	69%	80	78	75	75	67	75%
53	Qwen 3.5 35B	$0.016	2.6m	72%	76	76	76	74	71	75%
32	Grok 4.3	$0.012	8.3s	69%	78	77	77	76	65	75%
40	Mistral Large	$0.051	42.1s	72%	77	74	74	74	73	75%
46	Z.AI GLM 4.5	$0.0088	46.9s	66%	82	78	74	70	67	74%
22	DeepSeek V4 Flash	$0.0028	14.4s	72%	77	76	75	72	72	74%
43	GPT-5.1	$0.035	24.2s	67%	82	75	74	71	69	74%
58	DeepSeek V3.1	$0.0082	1.1m	64%	82	79	72	70	68	74%
29	Mistral Large 3	$0.013	22.7s	72%	76	75	75	72	72	74%
34	Gemini 3.5 Flash (Reasoning, Minimal)	$0.027	6.8s	71%	77	77	76	70	69	74%
39	DeepSeek V4 Pro	$0.012	21.9s	70%	76	76	75	69	68	73%
118	Gemma 4 31B (Reasoning)	$0.0055	7.5m	57%	80	79	76	75	46	71%
63	GPT-4.1	$0.032	10.8s	61%	82	73	72	69	60	71%
59	MiniMax M2.5	$0.0040	53.0s	64%	78	74	73	71	59	71%
98	Z.AI GLM 4.5 Air	$0.0099	4.8m	62%	79	72	69	68	65	70%
88	Qwen 3.6 35B	$0.017	1.3m	55%	82	79	78	69	42	70%
66	DeepSeek-V2 Chat	$0.0074	40.7s	62%	77	74	70	68	60	70%
81	GPT-5 Nano	$0.0093	2.6m	65%	74	72	71	66	62	69%
67	MiniMax M2.7	$0.0083	28.1s	61%	78	73	72	67	55	69%
87	Gemini 2.5 Flash Lite (Reasoning)	$0.0091	1.3m	56%	81	73	66	65	58	69%
62	Claude Haiku 4.5	$0.034	10.6s	65%	72	69	68	68	66	68%
117	Qwen 3.5 9B	$0.0044	7.6m	62%	74	70	69	69	58	68%
80	GPT-5.4 Nano (Reasoning, Low)	$0.0051	16.8s	56%	76	75	65	62	60	68%
120	Qwen3.7 Max	$0.056	4.5m	48%	86	84	79	46	42	67%
100	WizardLM 2 8x22b	$0.014	3.0m	56%	76	74	68	66	53	67%
79	Writer: Palmyra X5	$0.022	23.5s	59%	76	71	67	65	57	67%
92	Gemini 2.5 Flash (Reasoning)	$0.029	46.5s	53%	80	73	72	67	43	67%
78	Grok 4.20	$0.029	9.7s	60%	72	70	65	64	62	67%
68	Ministral 8B	$0.0024	22.0s	63%	69	68	66	65	63	66%
76	GPT-4.1 Mini	$0.0059	29.9s	61%	71	71	67	62	59	66%
72	GPT-4o, Aug. 6th (temp=0)	$0.044	9.6s	64%	67	66	66	65	64	65%
74	Grok 4.20 (Beta)	$0.029	6.8s	62%	67	67	64	64	64	65%
111	Gemma 4 26B (Reasoning)	$0.0044	3.7m	53%	80	76	76	47	46	65%
86	DeepSeek V3 (2024-12-26)	$0.0064	57.6s	58%	74	68	66	61	56	65%
77	Mistral Small 3.2 24B	$0.0023	17.3s	61%	68	67	67	67	56	65%
93	Claude 3 Haiku	$0.0079	14.6s	54%	71	70	62	57	53	62%
89	GPT-5.4 Mini	$0.012	9.7s	57%	68	63	61	61	59	62%
85	Gemini 2.5 Flash Lite	$0.0034	6.8s	57%	70	62	62	60	58	62%
94	GPT-4o, Aug. 6th (temp=1)	$0.039	11.1s	56%	68	64	61	60	58	62%
96	Mistral Medium 3.1	$0.016	33.0s	54%	68	64	59	58	58	61%
105	GPT-4o, May 13th (temp=1)	$0.121	6.7s	55%	67	66	62	61	51	61%
108	Llama 3.1 70B	$0.0087	45.7s	45%	74	72	68	63	28	61%
91	Ministral 3 8B	$0.0030	23.0s	58%	64	62	61	60	56	60%
116	Z.AI GLM 4.7 Flash	$0.0056	2.8m	49%	71	66	57	54	52	60%
109	GPT-4o, May 13th (temp=0)	$0.123	6.9s	54%	64	63	60	60	50	60%
112	DeepSeek V3 (2025-03-24)	$0.0049	1.1m	48%	76	62	62	46	46	59%
102	Mistral Small 4	$0.0043	9.2s	50%	69	59	58	52	50	57%
107	Qwen3 235B A22B Instruct 2507	$0.0025	1.3m	52%	63	60	59	57	46	57%
124	Qwen 3 32B	$0.0052	3.3m	47%	68	60	57	53	42	56%
110	Ministral 3 14B	$0.0046	39.7s	50%	62	54	54	54	52	55%
103	GPT-5.4 Nano	$0.0037	14.0s	53%	58	57	56	55	51	55%
114	Mistral Small 4 (Reasoning)	$0.0067	28.7s	46%	64	61	60	59	32	55%
123	Hermes 3 405B	$0.022	1.4m	44%	65	61	53	46	41	53%
122	Cydonia 24B V4.1	$0.0063	1.6m	45%	61	54	50	50	46	52%
127	Qwen 2.5 72B	$0.0080	1.6m	38%	62	56	45	43	40	49%
125	Gemma 3 27B	$0.0026	33.8s	43%	56	49	46	46	45	48%
119	Ministral 3B	$0.0009	13.4s	46%	51	48	48	47	45	48%
128	Nemotron 3 Nano	$0.0045	2.8m	43%	54	47	46	46	43	47%
121	Ministral 3 3B	$0.0011	14.4s	45%	51	48	48	44	42	47%
113	Nemotron 3 Super	$0.0000	3.2m	–	46	–	–	–	–	46%
131	Cohere Command R+ (Aug. 2024)	$0.085	1.5m	38%	55	49	46	46	29	45%
130	Arcee AI: Trinity Mini	$0.0069	3.8m	43%	47	46	45	44	40	45%
126	GPT-4.1 Nano	$0.0012	8.2s	37%	56	43	43	42	37	44%
129	Hermes 3 70B	$0.0065	56.2s	34%	55	46	42	35	27	41%
136	Rocinante 12B	$0.0046	35.8s	30%	46	46	36	33	28	38%
134	Gemma 3 12B	$0.0024	36.5s	32%	44	36	35	35	35	37%
133	Gemma 3 4B	$0.0009	34.5s	34%	40	39	36	35	30	36%
132	Mistral NeMO	$0.0035	19.1s	35%	38	37	37	37	32	36%
137	Skyfall 36B V2	$0.0072	28.5s	31%	46	37	36	34	27	36%
138	Llama 3.1 8B	$0.0005	1.3m	25%	44	38	29	28	27	33%
70.37%

Median	Evaluator	Top 3	Flop 3
74.1%	Alias accuracy	97Mistral Small 3.2 24B 95GPT-5.4 Mini (Reasoning, Low) 95GPT-5.4 Mini	0Nemotron 3 Super 0Gemma 3 12B 1Gemma 3 4B
100.0%	Character precision	100Grok 4.3 100Z.AI GLM 4.5 Air 100o4 Mini	0Gemma 3 12B 50Rocinante 12B 53Llama 3.1 8B
97.2%	Character recall	100Claude Opus 4.6 100Grok 4.20 (Beta) 100Gemini 3 Flash (Preview, Reasoning)	0Nemotron 3 Super 0Gemma 3 12B 7Gemma 3 4B
100.0%	Isolated character handling	100ByteDance Seed 2.0 Lite 100Qwen 3.5 35B 100Claude Opus 4	0Nemotron 3 Super 0Gemma 3 4B 1Skyfall 36B V2
100.0%	Red-herring resistance	100Qwen 3.5 Plus (2026-02-15) 100GPT-4.1 Mini 100GPT-4.1 Nano	73Hermes 3 70B 86Z.AI GLM 4.7 Flash 86Ministral 3 8B
9.1%	Relationship category recall	89GPT-5.4 (Reasoning) 80GPT-5.5 (Reasoning) 79Claude Opus 4.6 (Reasoning)	0Nemotron 3 Super 0Gemma 3 12B 0Nemotron 3 Nano
100.0%	Relationship endpoint integrity	100GPT-5.4 Nano 100Gemini 3.1 Pro (Preview) 100Mistral Large 2	0Gemma 3 12B 45Skyfall 36B V2 48Rocinante 12B
70.5%	Relationship precision	100Gemma 3 12B 100Gemini 3.1 Flash Lite (Preview) 100Nemotron 3 Super	2Grok 4.20 (Beta) 2Ministral 3 3B 2Gemma 3 4B
2.9%	Relationship recall	93GPT-5.4 (Reasoning) 83GPT-5.5 (Reasoning) 77Claude Opus 4.6 (Reasoning)	0Nemotron 3 Super 0Gemma 3 12B 0Nemotron 3 Nano
81.8%	Relationship type accuracy	100Gemma 3 12B 100Claude Opus 4.8 (Reasoning, Low) 100Nemotron 3 Super	23Mistral NeMO 31Gemma 3 27B 37Llama 3.1 8B
100.0%	XML structure	100GPT-5.4 Nano (Reasoning, Low) 100Claude Sonnet 4 100GPT-5.4 Nano (Reasoning)	10Nemotron 3 Super 40Rocinante 12B 45DeepSeek V3 (2025-03-24)