Short text (~524 words), small codex (11 entries)

Codex Red Herring (False Positive Detection)

Tests whether models correctly report "no violations" when a codex is fully consistent with the prose passage. Models that hallucinate false violations (false positives) fail. Uses a 2×2 matrix of text length × codex size, with bare and detailed-entry variants.

Hallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
Claude Opus 4.6 (Reasoning)	100%
Qwen3.7 Max	100%
Grok 4.5 (Reasoning, High)	100%
Gemini 3.1 Pro (Preview)	100%
Z.AI GLM 5.1	100%
Qwen3.6 Max Preview	100%
Gemini 3.5 Flash (Reasoning)	100%
Z.AI GLM 5 Turbo	100%
Claude Opus 4.7 (Reasoning)	100%
Claude Opus 4.8 (Reasoning)	100%
Claude Opus 4.6	100%
Claude Opus 4.8 (Reasoning, Low)	100%
GPT-5 Mini	100%
Grok 4.20 (Reasoning)	100%
GPT-5.6 Sol	100%
GPT-5.6 Luna (Reasoning)	100%
MiniMax M3	100%
Claude Opus 4.7	100%
GPT-5.4 Mini (Reasoning)	100%
Qwen 3.5 Plus (2026-04-20)	100%

	Score	Cost	Time
Gemma 4 26B	100%	$0.0002	3.8s
GPT-5.6 Luna	100%	$0.0005	709ms
GPT-5.4 Nano	80%	$0.0003	1.0s
GPT-5.4 Nano (Reasoning, Low)	100%	$0.0005	2.8s
GPT-5.6 Sol	100%	$0.0023	2.0s
Arcee AI: Trinity Mini	81%	$0.0012	36.6s
Gemma 4 31B	100%	$0.0002	12.4s
GPT-5.4 Nano (Reasoning)	100%	$0.0008	3.9s
ByteDance Seed 1.6 Flash	100%	$0.0004	5.3s
Inception Mercury 2	100%	$0.0013	2.0s
GPT-5.6 Luna (Reasoning)	100%	$0.0016	2.6s
GPT-5.5	100%	$0.0041	1.3s
Gemini 3.5 Flash (Reasoning, Minimal)	100%	$0.0024	850ms
Gemini 2.5 Flash Lite (Reasoning)	93%	$0.0012	8.6s
GPT-5.4 Mini (Reasoning, Low)	93%	$0.0019	3.6s
Mistral Small 4 (Reasoning)	100%	$0.0009	9.3s
Qwen 3 32B	78%	$0.0004	12.3s
Cohere Command R+ (Aug. 2024)	65%	$0.0046	2.5s
MiniMax M2.5	93%	$0.0015	12.4s
DeepSeek V4 Flash (Reasoning)	77%	$0.0004	17.6s

	Score	Cost	Speed	Stability
GPT-5.6 Luna	100%	$0.0005	709ms	100%
Gemma 4 26B	100%	$0.0002	3.8s	100%
GPT-5.4 Nano (Reasoning, Low)	100%	$0.0005	2.8s	100%
ByteDance Seed 1.6 Flash	100%	$0.0004	5.3s	100%
Inception Mercury 2	100%	$0.0013	2.0s	100%
GPT-5.4 Nano (Reasoning)	100%	$0.0008	3.9s	100%
GPT-5.6 Luna (Reasoning)	100%	$0.0016	2.6s	100%
Gemini 3.5 Flash (Reasoning, Minimal)	100%	$0.0024	850ms	100%
GPT-5.6 Sol	100%	$0.0023	2.0s	100%
Mistral Small 4 (Reasoning)	100%	$0.0009	9.3s	100%
Gemma 4 31B	100%	$0.0002	12.4s	100%
GPT-5.5	100%	$0.0041	1.3s	100%
Z.AI GLM 4.5 Air	100%	$0.0012	19.2s	100%
ByteDance Seed 1.6	100%	$0.0017	16.9s	100%
Z.AI GLM 5 Turbo	100%	$0.0036	9.8s	100%
GPT-5.4 Mini (Reasoning)	100%	$0.0039	9.0s	100%
MiniMax M2.7	100%	$0.0014	22.4s	100%
Gemini 2.5 Flash (Reasoning)	100%	$0.0051	8.5s	100%
Claude Opus 4.7	100%	$0.011	940ms	100%
Claude Opus 4.7 (Reasoning)	100%	$0.011	2.0s	100%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	# 6	# 7	# 8	# 9	# 10	Total
36	Claude Opus 4.6 (Reasoning)	$0.033	20.7s	100%	100	100	100	100	100	100	100	100	100	100	100%
35	Qwen3.7 Max	$0.026	47.2s	100%	100	100	100	100	100	100	100	100	100	100	100%
31	Grok 4.5 (Reasoning, High)	$0.020	41.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
38	Gemini 3.1 Pro (Preview)	$0.039	27.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
34	Z.AI GLM 5.1	$0.012	1.3m	100%	100	100	100	100	100	100	100	100	100	100	100%
46	Qwen3.6 Max Preview	$0.039	2.3m	100%	100	100	100	100	100	100	100	100	100	100	100%
37	Gemini 3.5 Flash (Reasoning)	$0.039	17.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
15	Z.AI GLM 5 Turbo	$0.0036	9.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
20	Claude Opus 4.7 (Reasoning)	$0.011	2.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
33	Claude Opus 4.8 (Reasoning)	$0.028	11.6s	100%	100	100	100	100	100	100	100	100	100	100	100%
29	Claude Opus 4.6	$0.022	9.7s	100%	100	100	100	100	100	100	100	100	100	100	100%
32	Claude Opus 4.8 (Reasoning, Low)	$0.028	10.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
21	GPT-5 Mini	$0.0044	29.1s	100%	100	100	100	100	100	100	100	100	100	100	100%
26	Grok 4.20 (Reasoning)	$0.0089	28.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
9	GPT-5.6 Sol	$0.0023	2.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
7	GPT-5.6 Luna (Reasoning)	$0.0016	2.6s	100%	100	100	100	100	100	100	100	100	100	100	100%
27	MiniMax M3	$0.0029	57.2s	100%	100	100	100	100	100	100	100	100	100	100	100%
19	Claude Opus 4.7	$0.011	940ms	100%	100	100	100	100	100	100	100	100	100	100	100%
16	GPT-5.4 Mini (Reasoning)	$0.0039	9.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
30	Qwen 3.5 Plus (2026-04-20)	$0.010	1.0m	100%	100	100	100	100	100	100	100	100	100	100	100%
28	Gemma 4 31B (Reasoning)	$0.0011	1.4m	100%	100	100	100	100	100	100	100	100	100	100	100%
14	ByteDance Seed 1.6	$0.0017	16.9s	100%	100	100	100	100	100	100	100	100	100	100	100%
12	GPT-5.5	$0.0041	1.3s	100%	100	100	100	100	100	100	100	100	100	100	100%
25	Gemma 4 26B (Reasoning)	$0.0012	52.7s	100%	100	100	100	100	100	100	100	100	100	100	100%
24	o4 Mini	$0.0097	18.9s	100%	100	100	100	100	100	100	100	100	100	100	100%
17	MiniMax M2.7	$0.0014	22.4s	100%	100	100	100	100	100	100	100	100	100	100	100%
8	Gemini 3.5 Flash (Reasoning, Minimal)	$0.0024	850ms	100%	100	100	100	100	100	100	100	100	100	100	100%
11	Gemma 4 31B	$0.0002	12.4s	100%	100	100	100	100	100	100	100	100	100	100	100%
1	GPT-5.6 Luna	$0.0005	709ms	100%	100	100	100	100	100	100	100	100	100	100	100%
2	Gemma 4 26B	$0.0002	3.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
18	Gemini 2.5 Flash (Reasoning)	$0.0051	8.5s	100%	100	100	100	100	100	100	100	100	100	100	100%
23	Z.AI GLM 4.7 Flash	$0.0012	47.1s	100%	100	100	100	100	100	100	100	100	100	100	100%
5	Inception Mercury 2	$0.0013	2.0s	100%	100	100	100	100	100	100	100	100	100	100	100%
22	Nemotron 3 Super	$0.0000	47.1s	100%	100	100	100	100	100	100	100	100	100	100	100%
13	Z.AI GLM 4.5 Air	$0.0012	19.2s	100%	100	100	100	100	100	100	100	100	100	100	100%
6	GPT-5.4 Nano (Reasoning)	$0.0008	3.9s	100%	100	100	100	100	100	100	100	100	100	100	100%
10	Mistral Small 4 (Reasoning)	$0.0009	9.3s	100%	100	100	100	100	100	100	100	100	100	100	100%
3	GPT-5.4 Nano (Reasoning, Low)	$0.0005	2.8s	100%	100	100	100	100	100	100	100	100	100	100	100%
4	ByteDance Seed 1.6 Flash	$0.0004	5.3s	100%	100	100	100	100	100	100	100	100	100	100	100%
42	Z.AI GLM 4.6	$0.0029	18.0s	56%	100	100	100	100	100	100	100	100	100	27	93%
43	Z.AI GLM 5.2 (Reasoning, High)	$0.0053	20.0s	55%	100	100	100	100	100	100	100	100	100	25	93%
45	GPT-5.1	$0.018	22.3s	55%	100	100	100	100	100	100	100	100	100	25	93%
53	Z.AI GLM 5	$0.013	1.4m	55%	100	100	100	100	100	100	100	100	100	25	93%
44	Qwen 3.6 Flash	$0.011	34.9s	55%	100	100	100	100	100	100	100	100	100	25	93%
60	Aion 3.0	$0.026	1.3m	55%	100	100	100	100	100	100	100	100	100	25	93%
49	o4 Mini High	$0.019	35.0s	55%	100	100	100	100	100	100	100	100	100	25	93%
41	MiniMax M2.5	$0.0015	12.4s	55%	100	100	100	100	100	100	100	100	100	25	93%
50	ByteDance Seed 2.0 Mini	$0.0018	1.8m	55%	100	100	100	100	100	100	100	100	100	25	93%
39	GPT-5.4 Mini (Reasoning, Low)	$0.0019	3.6s	55%	100	100	100	100	100	100	100	100	100	25	93%
40	Gemini 2.5 Flash Lite (Reasoning)	$0.0012	8.6s	55%	100	100	100	100	100	100	100	100	100	25	93%
48	Nemotron 3 Nano	$0.0016	1.7m	55%	100	100	100	100	100	100	100	100	100	25	93%
99	Claude Sonnet 4.6 (Reasoning)	$0.060	54.9s	40%	100	100	100	100	100	100	100	100	25	25	85%
51	Grok 4.5 (Reasoning, Low)	$0.0075	14.5s	40%	100	100	100	100	100	100	100	100	25	25	85%
59	Claude Sonnet 5 (Reasoning)	$0.018	18.3s	40%	100	100	100	100	100	100	100	100	25	25	85%
78	Qwen 3.5 27B	$0.028	2.3m	40%	100	100	100	100	100	100	100	100	25	25	85%
52	GPT-OSS 120B	$0.0006	49.8s	40%	100	100	100	100	100	100	100	100	25	25	85%
57	GPT-5 Nano	$0.0036	1.2m	40%	100	100	100	100	100	100	100	100	25	25	85%
56	Grok 4.3 (Reasoning)	$0.0073	41.2s	37%	100	100	100	100	100	100	100	100	25	17	84%
61	Arcee AI: Trinity Mini	$0.0012	36.6s	22%	100	100	100	100	100	100	100	100	6	0	81%
47	GPT-5.4 Nano	$0.0003	1.0s	38%	100	100	100	100	100	100	100	50	25	25	80%
55	Z.AI GLM 4.5	$0.0018	13.0s	32%	100	100	100	100	100	100	100	29	25	25	78%
63	Qwen 3.6 35B	$0.0080	47.0s	31%	100	100	100	100	100	100	100	25	25	25	78%
64	Aion 2.0	$0.0054	59.6s	31%	100	100	100	100	100	100	100	25	25	25	78%
54	Qwen 3 32B	$0.0004	12.3s	31%	100	100	100	100	100	100	100	25	25	25	78%
58	DeepSeek V4 Flash (Reasoning)	$0.0004	17.6s	29%	100	100	100	100	100	100	100	25	25	17	77%
62	GPT-5.2	$0.0091	10.8s	27%	100	100	100	100	100	100	100	25	25	13	76%
93	Qwen 3.6 27B	$0.025	1.8m	27%	100	100	100	100	100	100	100	25	25	13	76%
68	Aion 3.0 Mini	$0.0045	1.2m	29%	100	100	100	100	100	50	50	50	38	17	70%
146	Qwen 3.5 122B	$0.086	3.8m	27%	100	100	100	100	100	100	25	25	25	25	70%
69	Claude Sonnet 5 (Reasoning, Low)	$0.017	17.1s	27%	100	100	100	100	100	100	25	25	25	25	70%
70	Xiaomi MIMO v2.5 Pro	$0.011	53.1s	27%	100	100	100	100	100	100	25	25	25	25	70%
121	Qwen 3.5 9B	$0.0025	3.4m	21%	100	100	100	100	50	50	50	50	50	17	67%
67	Cohere Command R+ (Aug. 2024)	$0.0046	2.5s	14%	100	100	100	100	100	100	25	10	8	8	65%
141	Gemini 2.5 Pro	$0.072	54.5s	21%	100	100	100	100	100	50	25	25	25	25	65%
66	GPT-4.1 Nano	$0.0003	3.7s	18%	100	100	100	100	100	31	29	28	27	26	64%
145	MoonshotAI: Kimi K2.6	$0.037	5.6m	16%	100	100	100	100	100	25	25	25	25	25	63%
72	Xiaomi MIMO v2.5	$0.0050	22.9s	16%	100	100	100	100	100	25	25	25	25	25	63%
143	Qwen 3.5 397B A17B	$0.022	4.2m	14%	100	100	100	100	100	25	25	25	25	17	62%
122	Gemini 3 Flash (Preview, Reasoning)	$0.030	54.0s	14%	100	100	100	100	100	25	25	25	25	17	62%
74	GPT-5.4 (Reasoning, Low)	$0.0085	7.6s	13%	100	100	100	100	100	25	25	25	17	17	61%
65	Claude Haiku 4.5	$0.0029	3.2s	30%	100	100	50	50	50	50	50	50	50	50	60%
79	Llama 3.1 70B	$0.0032	38.5s	11%	100	100	100	100	100	35	30	17	5	4	59%
71	Qwen 2.5 72B	$0.0003	5.9s	11%	100	100	100	100	42	33	32	29	25	17	58%
75	Claude Sonnet 4	$0.0056	2.1s	10%	100	100	100	100	50	25	25	25	25	17	57%
73	GPT-4.1 Mini	$0.0009	3.8s	9%	100	100	100	100	35	33	31	31	30	7	57%
86	Hermes 3 405B	$0.0016	2.9s	3%	100	100	100	100	17	17	17	17	13	8	49%
76	Hermes 3 70B	$0.0005	5.9s	11%	100	100	100	50	42	38	33	13	7	5	49%
81	Cydonia 24B V4.1	$0.0004	6.3s	9%	100	100	100	50	35	29	28	17	10	8	48%
115	GPT-5.5 (Reasoning, Low)	$0.017	10.5s	8%	100	100	100	25	25	25	25	25	25	25	48%
136	Z.AI GLM 4.7	$0.014	2.0m	8%	100	100	100	25	25	25	25	25	25	25	48%
84	GPT-4.1	$0.0028	2.1s	12%	100	100	42	42	38	25	17	17	17	17	41%
144	GPT-5	$0.048	1.6m	10%	100	100	25	25	25	25	25	25	25	25	40%
87	GPT-5.6 Terra	$0.0016	1.4s	10%	100	100	25	25	25	25	25	25	25	25	40%
102	Claude Sonnet 4.5	$0.0069	3.5s	10%	100	100	42	25	25	25	25	25	17	17	40%
88	DeepSeek-V2 Chat	$0.0006	2.5s	10%	100	100	25	25	25	25	25	25	25	17	39%
89	GPT-5.4 Mini	$0.0007	3.4s	10%	100	100	25	25	25	25	25	25	25	17	39%
82	Claude Sonnet 4.6	$0.0099	6.2s	30%	42	38	38	38	38	35	35	35	35	8	34%
112	Grok 4.3	$0.0022	2.3s	5%	100	100	38	17	17	17	17	13	13	6	33%
80	Ministral 3 14B	$0.0004	5.3s	24%	50	50	30	30	30	29	29	29	29	28	33%
139	GPT-5.5 (Reasoning)	$0.041	19.5s	14%	100	25	25	25	25	25	25	25	25	25	33%
130	DeepSeek V4 Pro (Reasoning)	$0.0053	1.7m	14%	100	25	25	25	25	25	25	25	25	25	33%
138	Qwen 3.5 35B	$0.023	1.2m	14%	100	25	25	25	25	25	25	25	25	25	33%
77	Mistral Medium 3.1	$0.0011	2.8s	27%	42	42	42	35	32	31	30	30	25	17	32%
95	DeepSeek V3 (2024-12-26)	$0.0006	3.5s	14%	100	28	25	25	25	25	25	25	25	17	32%
100	GPT-5.4	$0.0026	1.7s	14%	100	25	25	25	25	25	25	25	25	17	32%
101	DeepSeek V4 Pro	$0.0022	31.9s	22%	50	50	50	35	31	30	25	17	13	13	31%
98	Grok 4.20	$0.0014	1.3s	13%	100	25	25	25	25	25	25	25	25	13	31%
106	GPT-4o, Aug. 6th (temp=1)	$0.0044	2.5s	13%	100	38	35	35	35	17	17	13	13	10	31%
134	GPT-5.4 (Reasoning)	$0.022	36.4s	13%	100	25	25	25	25	25	17	17	17	17	29%
83	Ministral 3B	$0.0001	3.5s	23%	50	30	30	29	28	28	28	28	27	8	29%
85	Ministral 3 3B	$0.0002	3.4s	24%	31	31	29	29	29	29	29	28	27	3	26%
90	Mistral Large 3	$0.0015	8.4s	25%	32	32	32	31	31	31	31	30	6	6	26%
104	DeepSeek V3.2	$0.0006	22.6s	21%	50	29	29	28	28	25	25	17	17	7	25%
120	GPT-4o Mini (temp=1)	$0.0003	3.7s	6%	100	33	32	31	13	13	10	8	7	7	25%
114	WizardLM 2 8x22b	$0.0011	26.8s	18%	50	50	33	25	25	25	17	13	8	6	25%
91	DeepSeek V4 Flash	$0.0002	6.9s	23%	33	32	31	30	29	28	25	25	10	7	25%
142	MoonshotAI: Kimi K2.5	$0.016	2.9m	25%	25	25	25	25	25	25	25	25	25	25	25%
107	Claude Opus 4.5	$0.0099	3.0s	25%	25	25	25	25	25	25	25	25	25	25	25%
97	Claude Sonnet 5	$0.0053	3.9s	25%	25	25	25	25	25	25	25	25	25	25	25%
125	Qwen 3.5 Flash	$0.0046	1.4m	25%	25	25	25	25	25	25	25	25	25	25	25%
119	ByteDance Seed 2.0 Lite	$0.0047	51.0s	25%	25	25	25	25	25	25	25	25	25	25	25%
92	Ministral 8B	$0.0003	7.2s	23%	28	28	28	27	27	27	27	27	26	2	25%
124	GPT-5.6 Sol (Reasoning)	$0.021	10.1s	24%	25	25	25	25	25	25	25	25	25	17	24%
111	GPT-5.6 Terra (Reasoning)	$0.0090	6.4s	24%	25	25	25	25	25	25	25	25	25	17	24%
105	Writer: Palmyra X5	$0.0025	12.5s	22%	29	28	25	25	25	25	25	17	17	13	23%
108	Mistral Large 2	$0.0059	7.8s	23%	35	33	31	31	31	30	10	10	8	7	23%
94	Gemini 2.5 Flash	$0.0006	875ms	23%	25	25	25	25	25	25	25	17	17	17	23%
96	DeepSeek V3 (2025-03-24)	$0.0005	5.7s	23%	25	25	25	25	25	25	25	17	17	17	23%
103	GPT-4o, Aug. 6th (temp=0)	$0.0032	1.3s	23%	25	25	25	25	25	25	25	17	17	13	22%
113	Qwen3 235B A22B Instruct 2507	$0.0002	4.6s	15%	30	25	25	25	17	17	17	17	17	17	20%
137	Qwen 3.5 Plus (2026-02-15)	$0.010	1.1m	13%	50	32	25	25	25	10	10	10	8	5	20%
110	Gemini 3.1 Flash Lite	$0.0007	1.3s	18%	35	25	25	25	25	17	13	13	13	10	20%
109	Gemini 3.1 Flash Lite (Reasoning)	$0.0006	1.4s	18%	25	25	25	25	25	17	17	13	13	13	20%
129	Ministral 3 8B	$0.0022	1.4m	20%	28	28	28	27	27	27	26	2	0	0	19%
123	GPT-4o Mini (temp=0)	$0.0003	3.8s	8%	35	35	35	31	10	10	10	8	8	8	19%
140	Claude Opus 4	$0.034	5.8s	15%	25	25	25	17	17	17	17	17	13	13	18%
116	DeepSeek V3.1	$0.0006	9.2s	17%	29	25	25	25	25	17	10	8	7	6	18%
118	Gemini 3 Flash (Preview)	$0.0012	1.6s	16%	25	17	17	17	17	17	17	17	13	13	17%
117	Gemini 3.1 Flash Lite (Preview)	$0.0007	1.2s	15%	25	25	17	17	17	17	13	13	13	13	17%
126	Mistral Small 4	$0.0005	4.8s	6%	30	29	28	28	8	7	6	6	5	5	15%
127	Gemini 2.5 Flash Lite	$0.0003	1.1s	9%	17	17	13	10	10	10	6	6	4	3	9%
128	Gemma 3 27B	$0.0002	8.7s	8%	13	10	10	8	8	8	7	6	6	6	8%
132	Mistral Small 3.2 24B	$0.0003	2.6s	5%	10	10	7	6	6	6	5	4	4	4	6%
133	Gemma 3 4B	$0.0001	4.8s	5%	7	7	7	6	6	6	6	6	6	4	6%
131	Mistral NeMO	$0.0003	2.8s	6%	8	7	7	7	6	6	5	5	4	3	6%
135	Gemma 3 12B	$0.0002	10.5s	5%	7	6	6	5	5	5	5	5	4	3	5%
62.18%

Median	Evaluator	Top 3	Flop 3
60.0%	Correct "no violations" response	100Claude Opus 4.7 100GPT-5.4 Nano (Reasoning) 100Gemma 4 26B (Reasoning)	0DeepSeek V3 (2025-03-24) 0GPT-4o, Aug. 6th (temp=0) 0Gemma 3 4B
77.5%	No hallucinated violations	100Nemotron 3 Super 100MiniMax M2.7 100Gemma 4 26B (Reasoning)	3Ministral 3 8B 4Ministral 8B 8Ministral 3 3B

Codex Red Herring (False Positive Detection)

Short text (~524 words), small codex (11 entries)

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)