Knowledge & reasoning

HealthBench (raw) leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on HealthBench (raw) is 69.4% by Claude Sonnet 5.5. 192 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Sonnet 5.5	69.4%	measured
2	Claude Opus 5.5	68.1%	measured
3	Claude Opus 5	67.1%	measured
4	Claude Fable 5.1	64.9%	estimated ± 6.4 pp, low confidence
5	Gemini 4 Argon	63.4%	estimated ± 6.4 pp, low confidence
6	Claude Fable 5	62.2%	estimated ± 6.4 pp, low confidence
7	GPT-5.6 Sol	57.4%	estimated ± 6.4 pp, low confidence
8	MiMo-V2.6-Pro	57.3%	estimated ± 6.4 pp, low confidence
9	GPT-6 Astra	56.9%	measured
10	Claude Opus 4.8	56.7%	estimated ± 6.4 pp, low confidence
11	Muse Spark 1.3	56.7%	estimated ± 6.4 pp, low confidence
12	GPT-6.1 Sol	56.7%	measured
13	Gemini 3.7 Flash	56.1%	estimated ± 6.4 pp, low confidence
14	Gemini 3.8 Flash	56.0%	estimated ± 6.4 pp, low confidence
15	Gemini 3.1 Pro	55.3%	estimated ± 6.4 pp, low confidence
16	Kimi K3	55.2%	estimated ± 6.4 pp, low confidence
17	Step 5 Preview	54.9%	estimated ± 6.4 pp, low confidence
18	Muse Spark 1.1	54.6%	estimated ± 6.4 pp, low confidence
19	GPT-5.5	54.3%	estimated ± 6.4 pp, low confidence
20	Muse Spark 1.2	54.0%	estimated ± 6.4 pp, low confidence
21	Claude Haiku 5.5	53.0%	estimated ± 6.4 pp, low confidence
22	GPT-5.4	52.4%	estimated ± 6.4 pp, low confidence
23	Grok 4.7	51.9%	estimated ± 6.4 pp, low confidence
24	Qwen3.8 Max Preview	51.9%	estimated ± 6.4 pp, low confidence
25	GPT-5.6 Terra	51.7%	estimated ± 6.4 pp, low confidence
26	Grok 4.6	51.7%	estimated ± 6.4 pp, low confidence
27	Gemini 3.5 Flash	51.5%	estimated ± 6.4 pp, low confidence
28	Grok 4.5	51.5%	estimated ± 6.4 pp, low confidence
29	GPT-5.3 Codex	51.4%	estimated ± 6.4 pp, low confidence
30	Claude Opus 4.7 (Adaptive)	51.2%	estimated ± 6.4 pp, low confidence
31	GLM-5.3	51.2%	estimated ± 6.4 pp, low confidence
32	Claude Sonnet 5	50.3%	estimated ± 6.4 pp, low confidence
33	GLM-5.2	50.1%	estimated ± 6.4 pp, low confidence
34	DeepSeek V4 Pro 0813	50.0%	estimated ± 6.4 pp, low confidence
35	GPT-6 Luna	50.0%	measured
36	Gemini 3.6 Flash	49.8%	estimated ± 6.4 pp, low confidence
37	Muse Spark	49.7%	estimated ± 6.4 pp, low confidence
38	Qwen3.7 Max	49.5%	estimated ± 6.4 pp, low confidence
39	Claude Opus 4.6 (Adaptive)	49.0%	estimated ± 6.4 pp, low confidence
40	GLM-5.3-Flash	49.0%	estimated ± 6.4 pp, low confidence
41	Gemini 3 Pro	48.8%	estimated ± 6.4 pp, low confidence
42	GPT-5.6 Luna	48.6%	estimated ± 6.4 pp, low confidence
43	Ling 3.1 Flash	48.5%	estimated ± 6.4 pp, low confidence
44	DeepSeek V4.1 Flash	48.3%	estimated ± 6.4 pp, low confidence
45	MiniMax M3	48.1%	estimated ± 6.4 pp, low confidence
46	DeepSeek V4 Flash 0731	47.8%	estimated ± 6.4 pp, low confidence
47	Qwen3.8-Flash-Next	47.2%	estimated ± 6.4 pp, low confidence
48	GPT-6 Sol	47.1%	measured
49	GPT-5.2	46.9%	estimated ± 6.4 pp, low confidence
50	Kimi K2.6	46.7%	estimated ± 6.4 pp, low confidence
51	Grok 4.3	46.4%	estimated ± 6.4 pp, low confidence
52	GPT-5.2-Codex	45.0%	estimated ± 6.4 pp, low confidence
53	MiMo-V2.5-Pro	45.0%	estimated ± 6.4 pp, low confidence
54	Qwen3.7 Plus	44.9%	estimated ± 6.4 pp, low confidence
55	MiMo-V2.6-Flash	44.4%	estimated ± 6.4 pp, low confidence
56	Kimi K2.7 Code	44.3%	estimated ± 6.4 pp, low confidence
57	Mistral Large 4	44.3%	estimated ± 6.4 pp, low confidence
58	Apodex 1.1	43.4%	estimated ± 6.4 pp, low confidence
59	Apodex 1.1 Mini	43.4%	estimated ± 6.4 pp, low confidence
60	Qwen3.8-27B	43.2%	estimated ± 6.4 pp, low confidence
61	Hy3	42.8%	estimated ± 6.4 pp, low confidence
62	Hy3 Preview	42.8%	estimated ± 6.4 pp, low confidence
63	Claude Opus 4.7	42.6%	estimated ± 6.4 pp, low confidence
64	Inkling-Small	42.6%	estimated ± 6.4 pp, low confidence
65	Inkling	41.2%	estimated ± 6.4 pp, low confidence
66	Qwen 3.6 Max (preview)	40.1%	estimated ± 6.4 pp, low confidence
67	Kimi K2.5	39.9%	estimated ± 6.4 pp, low confidence
68	Kimi K2.5 (Reasoning)	39.9%	estimated ± 6.4 pp, low confidence
69	MiMo-V2-Pro	39.6%	estimated ± 6.4 pp, low confidence
70	Claude Opus 4.5 Thinking	39.3%	estimated ± 6.4 pp, low confidence
71	GLM-5.1	39.3%	estimated ± 6.4 pp, low confidence
72	A.X K2	38.8%	estimated ± 6.4 pp, low confidence
73	MiniMax M2.7	38.8%	estimated ± 6.4 pp, low confidence
74	GLM-5	38.5%	estimated ± 6.4 pp, low confidence
75	Solar Pro 4	38.4%	estimated ± 6.4 pp, low confidence
76	GPT-5.1	37.6%	estimated ± 6.4 pp, low confidence
77	GPT-5 (high)	37.6%	estimated ± 6.4 pp, low confidence
78	Nemotron 3 Ultra	37.5%	estimated ± 6.4 pp, low confidence
79	GPT-5.4 nano	37.4%	estimated ± 6.4 pp, low confidence
80	GPT-5.4 mini	37.2%	estimated ± 6.4 pp, low confidence
81	GLM-5-Turbo	36.9%	estimated ± 6.4 pp, low confidence
82	Qwen3.6 Plus	36.9%	estimated ± 6.4 pp, low confidence
83	GLM-4.7	36.4%	estimated ± 6.4 pp, low confidence
84	Grok 4	35.7%	estimated ± 6.4 pp, low confidence
85	GPT-5.1-Codex	34.6%	estimated ± 6.4 pp, low confidence
86	GPT-5.1-Codex-Max	34.6%	estimated ± 6.4 pp, low confidence
87	GPT-5 (medium)	34.2%	estimated ± 6.4 pp, low confidence
88	Qwen3.5-122B-A10B	34.0%	estimated ± 6.4 pp, low confidence
89	Qwen3.5-27B	32.5%	estimated ± 6.4 pp, low confidence
90	Ling 3.0 Flash	32.3%	estimated ± 6.4 pp, low confidence
91	Ling 3.0 Flash FP8	32.3%	estimated ± 6.4 pp, low confidence
92	Gemma 4 31B	32.2%	estimated ± 6.4 pp, low confidence
93	Qwen3.6-27B	31.6%	estimated ± 6.4 pp, low confidence
94	o3-pro	30.9%	estimated ± 6.7 pp, low confidence
95	Gemini 2.5 Pro	30.9%	estimated ± 6.4 pp, low confidence
96	Qwen3.6-35B-A3B	30.6%	estimated ± 6.4 pp, low confidence
97	MiMo-V2-Omni	30.5%	estimated ± 6.4 pp, low confidence
98	Ling 3.0 Flash VL	30.3%	estimated ± 6.4 pp, low confidence
99	Muse Glimmer 30B	30.3%	estimated ± 6.4 pp, low confidence
100	Step 3.7 Flash	29.6%	estimated ± 6.4 pp, low confidence
101	Qwen3.5-35B-A3B	29.2%	estimated ± 6.4 pp, low confidence
102	Nemotron 3 Super 100B	28.9%	estimated ± 6.4 pp, low confidence
103	o3	28.1%	estimated ± 6.4 pp, low confidence
104	Qwen3.5 397B	27.7%	estimated ± 6.4 pp, low confidence
105	Qwen3.5 397B (Reasoning)	27.7%	estimated ± 6.4 pp, low confidence
106	GPT-OSS 120B	27.5%	estimated ± 6.4 pp, low confidence
107	Gemma 4 26B A4B	27.1%	estimated ± 6.4 pp, low confidence
108	Grok 4.1 Fast (Reasoning)	27.1%	estimated ± 6.4 pp, low confidence
109	Claude 4.1 Opus	26.9%	estimated ± 6.7 pp, low confidence
110	Claude Opus 4.6	26.9%	estimated ± 6.4 pp, low confidence
111	Grok 4 Fast (Reasoning)	26.9%	estimated ± 6.4 pp, low confidence
112	Gemini 3.5 Flash-Lite	26.5%	estimated ± 6.4 pp, low confidence
113	Quasar 438B	26.4%	estimated ± 6.4 pp, low confidence
114	K-EXAONE 2.0	26.3%	estimated ± 6.4 pp, low confidence
115	GLM-5V-Turbo	24.4%	estimated ± 6.4 pp, low confidence
116	DeepSeek-R1	22.8%	estimated ± 6.4 pp, low confidence
117	Trinity-Large-Preview	22.8%	estimated ± 6.4 pp, low confidence
118	Trinity-Large-Thinking	22.8%	estimated ± 6.4 pp, low confidence
119	Gemma 4 12B	22.6%	estimated ± 6.4 pp, low confidence
120	Gemini 3 Flash	21.7%	estimated ± 6.4 pp, low confidence
121	DeepSeek V3.1 (Reasoning)	20.8%	estimated ± 6.4 pp, low confidence
122	K-Exaone	20.3%	estimated ± 6.4 pp, low confidence
123	Mistral Medium 3.5 128B	20.2%	estimated ± 6.4 pp, low confidence
124	Claude Sonnet 4.6	19.5%	estimated ± 6.4 pp, low confidence
125	Claude Opus 4.5	19.4%	estimated ± 6.4 pp, low confidence
126	o1-pro	18.8%	estimated ± 6.7 pp, low confidence
127	Claude 4.1 Opus Thinking	18.5%	estimated ± 6.4 pp, low confidence
128	Command A+	17.8%	estimated ± 6.4 pp, low confidence
129	Qwen3 Max	17.6%	estimated ± 6.4 pp, low confidence
130	Mercury 2.5	17.5%	estimated ± 6.4 pp, low confidence
131	o1-preview	17.4%	estimated ± 6.7 pp, low confidence
132	Nemotron 3 Nano 30B	17.0%	estimated ± 6.4 pp, low confidence
133	DeepSeek V3.2	16.7%	estimated ± 6.4 pp, low confidence
134	Granite 4.2 30B	16.7%	estimated ± 6.4 pp, low confidence
135	North Mini Code	16.6%	estimated ± 6.4 pp, low confidence
136	GPT-OSS 20B	16.4%	estimated ± 6.4 pp, low confidence
137	Sarvam 105B	16.4%	estimated ± 6.4 pp, low confidence
138	Nemotron 3.5 Lightning 30B A3B NVFP4	15.9%	estimated ± 6.4 pp, low confidence
139	Solar Pro 3	15.5%	estimated ± 6.4 pp, low confidence
140	Mistral Small 4	14.9%	estimated ± 6.4 pp, low confidence
141	Mistral Small 4 (Reasoning)	14.9%	estimated ± 6.4 pp, low confidence
142	Granite 4.2 8B	14.6%	estimated ± 6.4 pp, low confidence
143	Ling 3.0 Tiny	14.1%	estimated ± 6.4 pp, low confidence
144	MiniCPM5-2B	13.5%	estimated ± 6.4 pp, low confidence
145	MiMo-V2-Flash	13.1%	estimated ± 6.4 pp, low confidence
146	Grok Code Fast 1	12.2%	estimated ± 6.4 pp, low confidence
147	o3-mini	12.1%	estimated ± 6.4 pp, low confidence
148	Qwen3-Omni-30B-A3B-Thinking	11.5%	estimated ± 6.4 pp, low confidence
149	Sarvam 30B	11.5%	estimated ± 6.4 pp, low confidence
150	Kimi K2	11.4%	estimated ± 6.4 pp, low confidence
151	Nemotron Ultra 253B	11.4%	estimated ± 6.4 pp, low confidence
152	GLM-4.5-Air	10.8%	estimated ± 6.4 pp, low confidence
153	o1	10.8%	estimated ± 6.4 pp, low confidence
154	LFM2.5-8B-A1B	10.6%	estimated ± 6.4 pp, low confidence
155	Celeris-1	10.5%	estimated ± 6.4 pp, low confidence
156	DeepSeek V3.1	10.3%	estimated ± 6.4 pp, low confidence
157	Granite 4.2 3B	10.2%	estimated ± 6.4 pp, low confidence
158	Granite-4.0-H-350M	9.9%	estimated ± 6.4 pp, low confidence
159	Ling 2.6 Flash	9.7%	estimated ± 6.4 pp, low confidence
160	LFM2.5-2.6B	9.6%	estimated ± 6.4 pp, low confidence
161	Exaone 4.0 1.2B	8.9%	estimated ± 6.4 pp, low confidence
162	GLM-4.6	8.6%	estimated ± 6.4 pp, low confidence
163	Granite-4.0-350M	8.6%	estimated ± 6.4 pp, low confidence
164	Grok 4.1 Fast	8.0%	estimated ± 6.4 pp, low confidence
165	LFM2.5-VL-1.6B-Extract	8.0%	estimated ± 6.4 pp, low confidence
166	Exaone 4.0 32B	7.8%	estimated ± 6.4 pp, low confidence
167	GPT-4.1 mini	7.8%	estimated ± 6.4 pp, low confidence
168	Granite-4.0-H-1B	7.8%	estimated ± 6.4 pp, low confidence
169	Phi-4 Multimodal Instruct	7.8%	estimated ± 6.4 pp, low confidence
170	Llama 4 Maverick	7.7%	estimated ± 6.4 pp, low confidence
171	Gemma 4 E2B	7.5%	estimated ± 6.4 pp, low confidence
172	Nemotron 3 Nano Omni 30B A3B	7.5%	estimated ± 6.4 pp, low confidence
173	DeepSeek V3 0324	7.4%	estimated ± 6.4 pp, low confidence
174	Gemini 2.5 Flash	7.4%	estimated ± 6.4 pp, low confidence
175	DeepSeek R1 Distill Qwen 32B	7.2%	estimated ± 6.4 pp, low confidence
176	Gemini 1.5 Pro	7.2%	estimated ± 6.4 pp, low confidence
177	Qwen3-Omni-30B-A3B-Instruct	7.2%	estimated ± 6.4 pp, low confidence
178	Gemma 3 27B	6.9%	estimated ± 6.4 pp, low confidence
179	Claude 4 Sonnet	6.8%	estimated ± 6.4 pp, low confidence
180	Gemini 1.0 Pro	6.6%	estimated ± 6.4 pp, low confidence
181	GPT-4.1	6.6%	estimated ± 6.4 pp, low confidence
182	GPT-4o mini	6.6%	estimated ± 6.4 pp, low confidence
183	Mistral Large 3	6.6%	estimated ± 6.4 pp, low confidence
184	Claude 3 Haiku	6.5%	estimated ± 6.4 pp, low confidence
185	Mistral Medium 3	6.5%	estimated ± 6.4 pp, low confidence
186	Llama 3.1 405B	6.3%	estimated ± 6.4 pp, low confidence
187	Gemma 4 E4B	6.0%	estimated ± 6.4 pp, low confidence
188	GPT-4.1 nano	6.0%	estimated ± 6.4 pp, low confidence
189	Llama 4 Scout	6.0%	estimated ± 6.4 pp, low confidence
190	Phi-4	6.0%	estimated ± 6.4 pp, low confidence
191	Solar Pro 2	5.8%	estimated ± 6.4 pp, low confidence
192	Ultravox v0.6 Llama 3.3 70B	5.7%	estimated ± 6.4 pp, low confidence
193	Qwen2.5 Coder 32B Instruct	5.5%	estimated ± 6.4 pp, low confidence
194	Mistral Large 2	5.2%	estimated ± 6.4 pp, low confidence
195	Nova Pro	5.1%	estimated ± 6.4 pp, low confidence
196	GPT-4 Turbo	4.9%	estimated ± 6.4 pp, low confidence
197	DeepSeek V3	4.6%	estimated ± 6.4 pp, low confidence
198	Claude 3 Opus	4.5%	estimated ± 6.4 pp, low confidence
199	GPT-4o	3.8%	estimated ± 6.4 pp, low confidence

HealthBench (raw) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General