Knowledge & reasoning

HealthBench Professional leaderboard

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

As of 2026-10-07, the highest measured score on HealthBench Professional is 69.2% by Claude Sonnet 5.5. 226 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	69.2%	measured
2	Claude Opus 5.5	65.6%	measured
3	Ling 3.1 Flash	65.4%	measured
4	Gemini 4 Argon	64.9%	estimated ± 4.8 pp, medium confidence
5	GPT-6 Astra	64.7%	measured
6	GPT-6.1 Sol	64.2%	measured
7	MiMo-V2.6-Pro	61.1%	estimated ± 4.8 pp, medium confidence
8	GPT-6 Luna	60.8%	measured
9	GPT-6 Sol	60.8%	measured
10	Sakana Fugu	60.7%	estimated ± 4.8 pp, medium confidence
11	Sakana Fugu-Ultra	60.7%	estimated ± 4.8 pp, medium confidence
12	GPT-5.6 Sol	60.5%	measured
13	Muse Spark 1.3	60.2%	estimated ± 4.8 pp, medium confidence
14	Claude Fable 5.1	60.0%	estimated ± 1.3 pp, low confidence
15	Claude Fable 5	60.0%	estimated ± 1.3 pp, medium confidence
16	Gemini 3.1 Pro	60.0%	estimated ± 1.3 pp, medium confidence
17	Gemini 3.8 Flash	60.0%	estimated ± 1.3 pp, medium confidence
18	Gemini 3.7 Flash	60.0%	estimated ± 1.3 pp, medium confidence
19	Claude Opus 4.7	59.9%	estimated ± 1.3 pp, medium confidence
20	Claude Opus 4.8	59.9%	estimated ± 1.3 pp, medium confidence
21	Gemini 3.5 Flash	59.9%	estimated ± 1.3 pp, medium confidence
22	Grok 4.6	59.9%	estimated ± 1.3 pp, medium confidence
23	Gemini 3.6 Flash	59.9%	estimated ± 1.3 pp, medium confidence
24	Qwen3.7 Max	59.9%	estimated ± 1.3 pp, medium confidence
25	Grok 4.5	59.9%	estimated ± 1.3 pp, medium confidence
26	Claude Opus 5	59.8%	measured
27	Gemini 3 Flash	59.7%	estimated ± 1.3 pp, medium confidence
28	Qwen3.8 Max	59.7%	estimated ± 1.3 pp, medium confidence
29	Muse Spark 1.2	59.6%	estimated ± 1.3 pp, medium confidence
30	GPT-5.5	59.5%	estimated ± 1.3 pp, medium confidence
31	Kimi K3	59.5%	estimated ± 1.3 pp, medium confidence
32	Muse Spark 1.1	59.3%	measured
33	Qwen3.6 Plus	59.2%	estimated ± 1.3 pp, medium confidence
34	Kimi K2.6	59.1%	estimated ± 1.3 pp, medium confidence
35	Claude Sonnet 5	59.0%	estimated ± 1.3 pp, medium confidence
36	Claude Sonnet 4.6	58.8%	estimated ± 1.3 pp, medium confidence
37	Muse Spark	58.8%	estimated ± 1.3 pp, medium confidence
38	DeepSeek V4 Pro 0813	58.3%	estimated ± 1.3 pp, medium confidence
39	GLM-5.1	58.1%	estimated ± 1.3 pp, medium confidence
40	Claude Mythos 5	58.1%	estimated ± 4.8 pp, medium confidence
41	GLM-5.3	57.9%	estimated ± 1.3 pp, medium confidence
42	Qwen3.8 Max Preview	57.7%	estimated ± 4.8 pp, medium confidence
43	GPT-5.3 Codex	57.7%	estimated ± 4.8 pp, medium confidence
44	GPT-5.6 Terra	57.7%	measured
45	GLM-5.2	57.7%	estimated ± 1.3 pp, medium confidence
46	GPT-5.2-Codex	57.7%	estimated ± 4.8 pp, low confidence
47	Kimi K2.7 Code	57.7%	estimated ± 4.8 pp, low confidence
48	MiMo-V2.6-Flash	57.7%	estimated ± 4.8 pp, low confidence
49	Mistral Large 4	57.7%	estimated ± 4.8 pp, low confidence
50	Apodex 1.1	57.7%	estimated ± 4.8 pp, low confidence
51	Apodex 1.1 Mini	57.7%	estimated ± 4.8 pp, low confidence
52	Hy3	57.7%	estimated ± 4.8 pp, low confidence
53	Celeris-1	57.7%	estimated ± 4.8 pp, low confidence
54	Claude 3 Haiku	57.7%	estimated ± 4.8 pp, low confidence
55	Claude 3 Opus	57.7%	estimated ± 4.8 pp, low confidence
56	Claude 4.1 Opus Thinking	57.7%	estimated ± 4.8 pp, low confidence
57	Claude 4 Sonnet	57.7%	estimated ± 4.8 pp, low confidence
58	Command A+	57.7%	estimated ± 4.8 pp, low confidence
59	DeepSeek-R1	57.7%	estimated ± 4.8 pp, low confidence
60	DeepSeek R1 Distill Qwen 32B	57.7%	estimated ± 4.8 pp, low confidence
61	DeepSeek V3 0324	57.7%	estimated ± 4.8 pp, low confidence
62	DeepSeek V3.1	57.7%	estimated ± 4.8 pp, low confidence
63	DeepSeek V3.1 (Reasoning)	57.7%	estimated ± 4.8 pp, low confidence
64	DeepSeek V3.2	57.7%	estimated ± 4.8 pp, low confidence
65	Exaone 4.0 1.2B	57.7%	estimated ± 4.8 pp, low confidence
66	Exaone 4.0 32B	57.7%	estimated ± 4.8 pp, low confidence
67	Gemini 1.0 Pro	57.7%	estimated ± 4.8 pp, low confidence
68	Gemini 1.5 Pro	57.7%	estimated ± 4.8 pp, low confidence
69	Gemini 2.5 Flash	57.7%	estimated ± 4.8 pp, low confidence
70	Gemma 3 27B	57.7%	estimated ± 4.8 pp, low confidence
71	GLM-4.5-Air	57.7%	estimated ± 4.8 pp, low confidence
72	GLM-5-Turbo	57.7%	estimated ± 4.8 pp, low confidence
73	GLM-5V-Turbo	57.7%	estimated ± 4.8 pp, low confidence
74	GPT-4 Turbo	57.7%	estimated ± 4.8 pp, low confidence
75	GPT-4o	57.7%	estimated ± 4.8 pp, low confidence
76	GPT-4o mini	57.7%	estimated ± 4.8 pp, low confidence
77	GPT-5.1	57.7%	estimated ± 4.8 pp, low confidence
78	GPT-5.1-Codex	57.7%	estimated ± 4.8 pp, low confidence
79	GPT-5.1-Codex-Max	57.7%	estimated ± 4.8 pp, low confidence
80	GPT-5 (high)	57.7%	estimated ± 4.8 pp, low confidence
81	GPT-5 (medium)	57.7%	estimated ± 4.8 pp, low confidence
82	GPT-OSS 120B	57.7%	estimated ± 4.8 pp, low confidence
83	GPT-OSS 20B	57.7%	estimated ± 4.8 pp, low confidence
84	Granite-4.0-350M	57.7%	estimated ± 4.8 pp, low confidence
85	Granite-4.0-H-1B	57.7%	estimated ± 4.8 pp, low confidence
86	Granite-4.0-H-350M	57.7%	estimated ± 4.8 pp, low confidence
87	Grok 4	57.7%	estimated ± 4.8 pp, low confidence
88	Grok 4.1 Fast	57.7%	estimated ± 4.8 pp, low confidence
89	Grok 4.1 Fast (Reasoning)	57.7%	estimated ± 4.8 pp, low confidence
90	Grok 4 Fast (Reasoning)	57.7%	estimated ± 4.8 pp, low confidence
91	Grok Code Fast 1	57.7%	estimated ± 4.8 pp, low confidence
92	K-Exaone	57.7%	estimated ± 4.8 pp, low confidence
93	Kimi K2	57.7%	estimated ± 4.8 pp, low confidence
94	LFM2.5-2.6B	57.7%	estimated ± 4.8 pp, low confidence
95	LFM2.5-8B-A1B	57.7%	estimated ± 4.8 pp, low confidence
96	LFM2.5-VL-1.6B-Extract	57.7%	estimated ± 4.8 pp, low confidence
97	Ling 3.0 Flash VL	57.7%	estimated ± 4.8 pp, low confidence
98	Ling 3.0 Tiny	57.7%	estimated ± 4.8 pp, low confidence
99	Llama 3.1 405B	57.7%	estimated ± 4.8 pp, low confidence
100	Llama 4 Maverick	57.7%	estimated ± 4.8 pp, low confidence
101	Llama 4 Scout	57.7%	estimated ± 4.8 pp, low confidence
102	MiMo-V2-Omni	57.7%	estimated ± 4.8 pp, low confidence
103	MiMo-V2-Pro	57.7%	estimated ± 4.8 pp, low confidence
104	Mistral Large 2	57.7%	estimated ± 4.8 pp, low confidence
105	Mistral Large 3	57.7%	estimated ± 4.8 pp, low confidence
106	Mistral Medium 3	57.7%	estimated ± 4.8 pp, low confidence
107	Mistral Small 4	57.7%	estimated ± 4.8 pp, low confidence
108	Mistral Small 4 (Reasoning)	57.7%	estimated ± 4.8 pp, low confidence
109	Muse Glimmer 30B	57.7%	estimated ± 4.8 pp, low confidence
110	Nemotron 3 Nano 30B	57.7%	estimated ± 4.8 pp, low confidence
111	Nemotron 3 Super 100B	57.7%	estimated ± 4.8 pp, low confidence
112	Nemotron Ultra 253B	57.7%	estimated ± 4.8 pp, low confidence
113	North Mini Code	57.7%	estimated ± 4.8 pp, low confidence
114	Nova Pro	57.7%	estimated ± 4.8 pp, low confidence
115	o3	57.7%	estimated ± 4.8 pp, low confidence
116	Phi-4	57.7%	estimated ± 4.8 pp, low confidence
117	Phi-4 Multimodal Instruct	57.7%	estimated ± 4.8 pp, low confidence
118	Quasar 438B	57.7%	estimated ± 4.8 pp, low confidence
119	Qwen2.5 Coder 32B Instruct	57.7%	estimated ± 4.8 pp, low confidence
120	Qwen3.5 397B (Reasoning)	57.7%	estimated ± 4.8 pp, low confidence
121	Qwen 3.6 Max (preview)	57.7%	estimated ± 4.8 pp, low confidence
122	Qwen3 Max	57.7%	estimated ± 4.8 pp, low confidence
123	Qwen3-Omni-30B-A3B-Instruct	57.7%	estimated ± 4.8 pp, low confidence
124	Qwen3-Omni-30B-A3B-Thinking	57.7%	estimated ± 4.8 pp, low confidence
125	Sarvam 105B	57.7%	estimated ± 4.8 pp, low confidence
126	Sarvam 30B	57.7%	estimated ± 4.8 pp, low confidence
127	Solar Pro 2	57.7%	estimated ± 4.8 pp, low confidence
128	Solar Pro 3	57.7%	estimated ± 4.8 pp, low confidence
129	Step 3.7 Flash	57.7%	estimated ± 4.8 pp, low confidence
130	Ultravox v0.6 Llama 3.3 70B	57.7%	estimated ± 4.8 pp, low confidence
131	Pareto 26.9	57.0%	estimated ± 4.8 pp, medium confidence
132	Step 5 Preview	56.9%	estimated ± 4.8 pp, medium confidence
133	Claude Opus 4.7 (Adaptive)	56.8%	estimated ± 4.3 pp, medium confidence
134	GPT-5.5 Pro	56.8%	estimated ± 4.6 pp, high confidence
135	Grok 4.7	56.7%	measured
136	Grok 4.20	56.6%	estimated ± 1.3 pp, medium confidence
137	Inkling	56.6%	estimated ± 1.3 pp, medium confidence
138	GPT-5.4 Pro	56.4%	estimated ± 4.6 pp, high confidence
139	DeepSeek V4 Flash 0731	56.3%	estimated ± 1.3 pp, medium confidence
140	Gemini 3.1 Flash-Lite	56.3%	estimated ± 1.3 pp, medium confidence
141	GLM-5.3-Flash	56.0%	estimated ± 1.3 pp, medium confidence
142	dots3-note Preview	55.8%	estimated ± 4.6 pp, high confidence
143	GPT-5.6 Luna	55.7%	measured
144	Ornith-1.5-397B	55.6%	estimated ± 4.8 pp, medium confidence
145	Claude Opus 4.6 (Adaptive)	55.1%	estimated ± 4.6 pp, high confidence
146	Claude Opus 4.5 Thinking	55.1%	estimated ± 4.6 pp, medium confidence
147	Claude Sonnet 4.5	55.1%	estimated ± 4.6 pp, medium confidence
148	Claude Sonnet 4.5 Thinking	55.1%	estimated ± 4.6 pp, medium confidence
149	Gemini 3 Pro	55.1%	estimated ± 4.6 pp, medium confidence
150	Gemini 3 Pro Deep Think	55.1%	estimated ± 4.6 pp, medium confidence
151	GPT-5.2	55.1%	estimated ± 4.6 pp, medium confidence
152	o1-preview	55.1%	estimated ± 4.9 pp, low confidence
153	o3-pro	55.1%	estimated ± 4.9 pp, low confidence
154	Gemini 3.5 Flash-Lite	55.0%	estimated ± 1.3 pp, low confidence
155	Grok 4.3	55.0%	estimated ± 1.3 pp, low confidence
156	Nemotron 3 Ultra	55.0%	estimated ± 1.3 pp, low confidence
157	Pareto 26.10 Preview	54.8%	estimated ± 4.8 pp, medium confidence
158	Hy4 preview	54.6%	estimated ± 4.8 pp, medium confidence
159	Claude Haiku 5.5	54.5%	estimated ± 4.8 pp, medium confidence
160	Inkling-Small	54.4%	estimated ± 1.3 pp, low confidence
161	Qwen3.8-Flash-Next	53.5%	estimated ± 4.8 pp, low confidence
162	Claude Opus 4.6	52.7%	estimated ± 4.8 pp, low confidence
163	Qwen3.8-Omni-Flash	52.1%	estimated ± 4.8 pp, low confidence
164	DeepSeek V4.1 Flash	51.9%	estimated ± 4.8 pp, low confidence
165	GPT-5.4 mini	51.7%	estimated ± 1.3 pp, low confidence
166	MiMo-V2.5-Pro	51.7%	estimated ± 1.3 pp, low confidence
167	Qwen3.8-27B	51.2%	estimated ± 1.3 pp, low confidence
168	Beam	51.1%	estimated ± 4.8 pp, low confidence
169	MiniMax M3	51.0%	estimated ± 1.3 pp, low confidence
170	Qwen3.5 Flash	50.9%	estimated ± 1.3 pp, low confidence
171	Qwen3.7 Plus	50.7%	estimated ± 4.8 pp, low confidence
172	MiMo-V2.5	50.1%	estimated ± 1.3 pp, low confidence
173	GLM-4.7	50.0%	estimated ± 1.3 pp, low confidence
174	Interfaze Beta	50.0%	estimated ± 4.8 pp, low confidence
175	GLM-4.6	49.9%	estimated ± 1.3 pp, low confidence
176	Ling 3.0 Flash	49.9%	estimated ± 1.3 pp, low confidence
177	GLM-4.5	49.9%	estimated ± 1.3 pp, low confidence
178	MiniMax M2.7	49.8%	estimated ± 1.3 pp, low confidence
179	Claude Haiku 4.5	49.8%	estimated ± 1.3 pp, low confidence
180	GPT-5.4 nano	49.8%	estimated ± 1.3 pp, low confidence
181	Laguna M.1	49.8%	estimated ± 1.3 pp, low confidence
182	Laguna XS.2	49.8%	estimated ± 1.3 pp, low confidence
183	Mistral Medium 3.5 128B	49.8%	estimated ± 1.3 pp, low confidence
184	Ornith-1.5-35B-A3B	48.6%	estimated ± 4.8 pp, low confidence
185	Solar Pro 4	48.2%	estimated ± 4.8 pp, low confidence
186	GPT-5.4	48.1%	measured
187	Qwen3.5 397B	47.0%	estimated ± 4.8 pp, low confidence
188	Qwen3.6-27B	45.9%	estimated ± 4.8 pp, low confidence
189	Kimi K2.5	45.5%	estimated ± 4.8 pp, low confidence
190	Kimi K2.5 (Reasoning)	45.5%	estimated ± 4.8 pp, low confidence
191	Hy3 Preview	44.7%	estimated ± 4.8 pp, low confidence
192	Claude Opus 4.5	44.3%	estimated ± 4.8 pp, low confidence
193	Qwen3.5-122B-A10B	43.6%	estimated ± 4.8 pp, low confidence
194	Ornith-1.5-9B	43.2%	estimated ± 4.8 pp, low confidence
195	Solar Open 2	43.0%	estimated ± 4.8 pp, low confidence
196	GLM-5	42.4%	estimated ± 4.8 pp, low confidence
197	Qwen3.6-35B-A3B	42.4%	estimated ± 4.8 pp, low confidence
198	MiMo-V2-Flash	42.0%	estimated ± 1.7 pp, low confidence
199	Ternary Bonsai 2 27B	42.0%	estimated ± 4.8 pp, low confidence
200	A.X K2	41.7%	estimated ± 4.8 pp, low confidence
201	Qwen3.5-27B	41.5%	estimated ± 4.8 pp, low confidence
202	Gemma 4 31B	39.2%	estimated ± 4.8 pp, low confidence
203	MAI-Thinking-1	39.0%	estimated ± 4.8 pp, low confidence
204	Qwen3.5-35B-A3B	39.0%	estimated ± 4.8 pp, low confidence
205	Ling 3.0 Flash FP8	38.6%	estimated ± 4.8 pp, low confidence
206	Gemini 2.5 Pro	36.8%	estimated ± 4.8 pp, low confidence
207	K-EXAONE 2.0	35.3%	estimated ± 4.8 pp, low confidence
208	Mercury 2.5	29.7%	estimated ± 4.8 pp, low confidence
209	o1-pro	29.7%	estimated ± 4.8 pp, low confidence
210	Gemma 4 12B	29.3%	estimated ± 4.8 pp, low confidence
211	Qwen3 235B 2507	27.2%	estimated ± 4.8 pp, low confidence
212	o3-mini	26.7%	estimated ± 4.8 pp, low confidence
213	Trinity-Large-Thinking	25.3%	estimated ± 4.8 pp, low confidence
214	o1	24.3%	estimated ± 4.8 pp, low confidence
215	Nemotron 3.5 Lightning 30B A3B NVFP4	24.1%	estimated ± 4.8 pp, low confidence
216	Nemotron 3 Nano Omni 30B A3B	19.3%	estimated ± 4.8 pp, low confidence
217	ZAYA1-8B	17.7%	estimated ± 4.8 pp, low confidence
218	MiniCPM5-2B	16.7%	estimated ± 4.8 pp, low confidence
219	LongCat-Flash-Lite-Sparse	15.8%	estimated ± 4.8 pp, low confidence
220	Gemma 4 26B A4B	13.7%	estimated ± 4.8 pp, low confidence
221	Granite 4.2 30B	12.5%	estimated ± 4.8 pp, low confidence
222	GPT-4.1	12.4%	estimated ± 4.8 pp, low confidence
223	GPT-4.1 mini	10.4%	estimated ± 4.8 pp, low confidence
224	Granite 4.2 8B	10.3%	estimated ± 4.8 pp, low confidence
225	Trinity-Large-Preview	9.6%	estimated ± 4.8 pp, low confidence
226	Claude 3.5 Sonnet	6.7%	estimated ± 4.8 pp, low confidence
227	DeepSeek V3	6.5%	estimated ± 4.8 pp, low confidence
228	Ling 2.6 Flash	6.5%	estimated ± 4.8 pp, low confidence
229	Gemma 4 E4B	6.2%	estimated ± 4.8 pp, low confidence
230	Mellum2-12B-A2.5B-Thinking	5.6%	estimated ± 4.8 pp, low confidence
231	ZAYA1-74B-Preview	5.5%	estimated ± 4.8 pp, low confidence
232	Granite 4.2 3B	4.2%	estimated ± 4.8 pp, low confidence
233	GPT-4.1 nano	2.6%	estimated ± 4.8 pp, low confidence
234	LLaDA2.2-mini	1.2%	estimated ± 4.8 pp, low confidence
235	Gemma 4 E2B	1.1%	estimated ± 4.8 pp, low confidence
236	Soofi S 30B-A3B	1.1%	estimated ± 4.8 pp, low confidence
237	Mellum2-12B-A2.5B-Instruct	0.8%	estimated ± 4.8 pp, low confidence
238	MiniCPM5-1B	0.1%	estimated ± 4.8 pp, low confidence
239	LFM2.5-VL-450M	0.0%	estimated ± 4.8 pp, low confidence
240	LFM2.5-230M	0.0%	estimated ± 4.8 pp, low confidence

HealthBench Professional leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General