Knowledge & reasoning

HealthBench Hard leaderboard

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

As of 2026-10-07, the highest measured score on HealthBench Hard is 42.8% by Muse Spark. 213 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Muse Spark	42.8%	measured
2	GPT-5.4	40.1%	measured
3	Sakana Fugu	36.6%	estimated ± 5.1 pp, low confidence
4	Sakana Fugu-Ultra	36.6%	estimated ± 5.1 pp, low confidence
5	GPT-6 Astra	36.6%	measured
6	Claude Opus 4.7 (Adaptive)	36.5%	estimated ± 5.1 pp, low confidence
7	Claude Mythos 5	36.4%	estimated ± 5.1 pp, low confidence
8	GPT-6.1 Sol	36.2%	measured
9	Claude Opus 4.8	36.1%	estimated ± 5.1 pp, low confidence
10	GPT-5.5	36.1%	estimated ± 5.1 pp, low confidence
11	Kimi K3	36.0%	estimated ± 5.1 pp, low confidence
12	Claude Opus 5.5	35.2%	estimated ± 7.7 pp, low confidence
13	Claude Sonnet 5.5	35.1%	estimated ± 7.7 pp, low confidence
14	Claude Fable 5.1	35.0%	estimated ± 7.7 pp, low confidence
15	Gemini 4 Argon	35.0%	estimated ± 7.7 pp, low confidence
16	Claude Opus 5	34.9%	estimated ± 7.7 pp, low confidence
17	Claude Fable 5	34.8%	estimated ± 7.7 pp, low confidence
18	Muse Spark 1.3	34.7%	estimated ± 7.7 pp, low confidence
19	Grok 4.7	34.6%	estimated ± 7.7 pp, low confidence
20	MiMo-V2.6-Pro	34.6%	estimated ± 7.7 pp, low confidence
21	Qwen3.8 Max Preview	34.5%	estimated ± 7.7 pp, low confidence
22	Ornith-1.5-397B	34.4%	estimated ± 5.1 pp, low confidence
23	GLM-5.3	34.4%	estimated ± 7.7 pp, low confidence
24	Grok 4.6	34.3%	estimated ± 7.7 pp, low confidence
25	GPT-5.5 Pro	34.3%	estimated ± 9.0 pp, low confidence
26	Step 5 Preview	34.2%	estimated ± 7.7 pp, low confidence
27	GPT-5.4 Pro	34.2%	estimated ± 9.0 pp, low confidence
28	Claude Haiku 5.5	34.2%	estimated ± 7.7 pp, low confidence
29	GLM-5.3-Flash	33.9%	estimated ± 7.7 pp, low confidence
30	Ling 3.1 Flash	33.7%	estimated ± 7.7 pp, low confidence
31	Gemini 3 Pro Deep Think	33.7%	estimated ± 9.0 pp, low confidence
32	Gemini 3.8 Flash	33.7%	estimated ± 7.7 pp, low confidence
33	Qwen3.8 Max	33.4%	estimated ± 5.1 pp, low confidence
34	Muse Spark 1.2	33.3%	estimated ± 7.7 pp, low confidence
35	Gemini 3.7 Flash	33.2%	estimated ± 7.7 pp, low confidence
36	GPT-5.6 Sol	33.1%	measured
37	Grok 4.5	33.1%	estimated ± 7.7 pp, low confidence
38	Mistral Large 4	32.9%	estimated ± 7.7 pp, low confidence
39	Claude Sonnet 5	32.9%	estimated ± 7.7 pp, low confidence
40	MiMo-V2.6-Flash	32.8%	estimated ± 7.7 pp, low confidence
41	GPT-5.6 Terra	32.7%	measured
42	GPT-5.2	32.1%	estimated ± 5.1 pp, low confidence
43	Qwen3.7 Max	32.1%	estimated ± 5.1 pp, low confidence
44	GPT-5.6 Luna	32.0%	measured
45	GPT-6 Luna	31.4%	measured
46	Hy4 preview	31.2%	estimated ± 5.1 pp, low confidence
47	Gemini 3.6 Flash	30.7%	estimated ± 7.7 pp, low confidence
48	Muse Spark 1.1	30.5%	estimated ± 7.7 pp, low confidence
49	Gemini 3.5 Flash	30.2%	estimated ± 5.1 pp, low confidence
50	GPT-6 Sol	30.1%	measured
51	GPT-5.3 Codex	29.5%	estimated ± 7.7 pp, low confidence
52	Claude Opus 4.6 (Adaptive)	28.9%	estimated ± 7.7 pp, low confidence
53	Claude Opus 4.7	27.8%	estimated ± 7.7 pp, low confidence
54	MiniMax M3	25.6%	estimated ± 7.7 pp, low confidence
55	Claude Opus 4.5 Thinking	25.5%	estimated ± 7.7 pp, low confidence
56	MiMo-V2-Pro	24.8%	estimated ± 7.7 pp, low confidence
57	GPT-5.2-Codex	24.5%	estimated ± 7.7 pp, low confidence
58	Qwen 3.6 Max (preview)	24.3%	estimated ± 7.7 pp, low confidence
59	Solar Pro 4	24.0%	estimated ± 7.7 pp, low confidence
60	Gemini 3 Pro	23.7%	estimated ± 7.7 pp, low confidence
61	Qwen3.8-Flash-Next	23.3%	estimated ± 5.1 pp, low confidence
62	Quasar 438B	21.5%	estimated ± 7.7 pp, low confidence
63	GLM-5-Turbo	21.2%	estimated ± 7.7 pp, low confidence
64	Apodex 1.1	20.8%	estimated ± 7.7 pp, low confidence
65	Apodex 1.1 Mini	20.8%	estimated ± 7.7 pp, low confidence
66	Gemini 3.1 Pro	20.6%	measured
67	Grok 4.20	20.3%	measured
68	GLM-5.1	20.1%	estimated ± 7.7 pp, low confidence
69	MiMo-V2.5-Pro	20.0%	estimated ± 7.7 pp, low confidence
70	Kimi K2.7 Code	19.6%	estimated ± 7.7 pp, low confidence
71	Hy3	18.6%	estimated ± 7.7 pp, low confidence
72	GPT-5.1	17.4%	estimated ± 7.7 pp, low confidence
73	Ling 3.0 Flash VL	17.0%	estimated ± 7.7 pp, low confidence
74	MiMo-V2-Omni	15.6%	estimated ± 7.7 pp, low confidence
75	GPT-5.1-Codex	15.1%	estimated ± 7.7 pp, low confidence
76	GPT-5.1-Codex-Max	15.1%	estimated ± 7.7 pp, low confidence
77	Claude Opus 4.6	14.8%	measured
78	GLM-5V-Turbo	14.7%	estimated ± 7.7 pp, low confidence
79	GLM-5.2	14.3%	estimated ± 5.1 pp, low confidence
80	GPT-5 (high)	13.6%	estimated ± 7.7 pp, low confidence
81	GPT-5 (medium)	13.3%	estimated ± 7.7 pp, low confidence
82	Claude 4.1 Opus Thinking	13.3%	estimated ± 7.7 pp, low confidence
83	MiniMax M2.7	13.1%	estimated ± 7.7 pp, low confidence
84	Command A+	12.5%	estimated ± 7.7 pp, low confidence
85	Grok 4	12.5%	estimated ± 7.7 pp, low confidence
86	Gemini 3.5 Flash-Lite	11.8%	estimated ± 7.7 pp, low confidence
87	o3-pro	11.2%	estimated ± 7.7 pp, low confidence
88	Qwen3.8-Omni-Flash	11.0%	estimated ± 5.1 pp, low confidence
89	A.X K2	10.3%	estimated ± 7.7 pp, low confidence
90	Qwen3.5 397B (Reasoning)	10.3%	estimated ± 7.7 pp, low confidence
91	DeepSeek V4.1 Flash	9.5%	estimated ± 5.1 pp, low confidence
92	Grok 4.1 Fast (Reasoning)	8.2%	estimated ± 7.7 pp, low confidence
93	o3	7.9%	estimated ± 7.7 pp, low confidence
94	K-EXAONE 2.0	7.0%	estimated ± 7.7 pp, low confidence
95	Step 3.7 Flash	6.6%	estimated ± 7.7 pp, low confidence
96	Claude 4.1 Opus	5.2%	estimated ± 7.7 pp, low confidence
97	Kimi K2.6	5.0%	estimated ± 5.1 pp, low confidence
98	Grok 4 Fast (Reasoning)	4.4%	estimated ± 7.7 pp, low confidence
99	Gemini 3 Flash	4.3%	estimated ± 7.7 pp, low confidence
100	Qwen3.6 Plus	4.2%	estimated ± 5.1 pp, low confidence
101	Muse Glimmer 30B	3.8%	estimated ± 7.7 pp, low confidence
102	Qwen3.7 Plus	3.5%	estimated ± 5.1 pp, low confidence
103	Gemma 4 26B A4B	2.9%	estimated ± 7.7 pp, low confidence
104	Claude 4 Sonnet	2.9%	estimated ± 7.7 pp, low confidence
105	DeepSeek V4 Pro 0813	2.4%	estimated ± 5.1 pp, low confidence
106	Grok 4.3	2.4%	estimated ± 5.1 pp, low confidence
107	DeepSeek V3.2	2.4%	estimated ± 7.7 pp, low confidence
108	Qwen3 Max	2.0%	estimated ± 7.7 pp, low confidence
109	Claude Sonnet 4.6	1.7%	estimated ± 5.1 pp, low confidence
110	Interfaze Beta	1.7%	estimated ± 5.1 pp, low confidence
111	GLM-4.6	1.6%	estimated ± 7.7 pp, low confidence
112	K-Exaone	1.3%	estimated ± 7.7 pp, low confidence
113	Mistral Medium 3.5 128B	1.2%	estimated ± 7.7 pp, low confidence
114	Grok Code Fast 1	1.1%	estimated ± 7.7 pp, low confidence
115	DeepSeek V3.1	1.0%	estimated ± 7.7 pp, low confidence
116	DeepSeek V3.1 (Reasoning)	0.9%	estimated ± 7.7 pp, low confidence
117	Inkling-Small	0.8%	estimated ± 5.1 pp, low confidence
118	DeepSeek-R1	0.7%	estimated ± 7.7 pp, low confidence
119	Nemotron 3 Super 100B	0.7%	estimated ± 7.7 pp, low confidence
120	Kimi K2	0.6%	estimated ± 7.7 pp, low confidence
121	MiniCPM5-2B	0.5%	estimated ± 7.7 pp, low confidence
122	Mercury 2.5	0.5%	estimated ± 7.7 pp, low confidence
123	Ornith-1.5-35B-A3B	0.4%	estimated ± 5.1 pp, low confidence
124	Qwen3.8-27B	0.4%	estimated ± 5.1 pp, low confidence
125	GPT-OSS 120B	0.4%	estimated ± 7.7 pp, low confidence
126	o1-preview	0.3%	estimated ± 7.7 pp, low confidence
127	Grok 4.1 Fast	0.3%	estimated ± 7.7 pp, low confidence
128	Mistral Small 4	0.3%	estimated ± 7.7 pp, low confidence
129	Mistral Small 4 (Reasoning)	0.3%	estimated ± 7.7 pp, low confidence
130	GLM-4.5-Air	0.3%	estimated ± 7.7 pp, low confidence
131	Ling 3.0 Tiny	0.3%	estimated ± 7.7 pp, low confidence
132	Trinity-Large-Preview	0.2%	estimated ± 7.7 pp, low confidence
133	Trinity-Large-Thinking	0.2%	estimated ± 7.7 pp, low confidence
134	Llama 4 Maverick	0.1%	estimated ± 7.7 pp, low confidence
135	North Mini Code	0.1%	estimated ± 7.7 pp, low confidence
136	Gemini 2.5 Flash	0.1%	estimated ± 7.7 pp, low confidence
137	DeepSeek V3 0324	0.1%	estimated ± 7.7 pp, low confidence
138	Mistral Large 3	0.1%	estimated ± 7.7 pp, low confidence
139	Qwen3.5 397B	0.1%	estimated ± 5.1 pp, low confidence
140	Mistral Medium 3	0.1%	estimated ± 7.7 pp, low confidence
141	GPT-OSS 20B	0.1%	estimated ± 7.7 pp, low confidence
142	Nemotron 3 Nano 30B	0.1%	estimated ± 7.7 pp, low confidence
143	Sarvam 105B	0.1%	estimated ± 7.7 pp, low confidence
144	Claude 3 Opus	0.1%	estimated ± 7.7 pp, low confidence
145	GPT-4o	0.1%	estimated ± 7.7 pp, low confidence
146	LFM2.5-2.6B	0.1%	estimated ± 7.7 pp, low confidence
147	DeepSeek R1 Distill Qwen 32B	0.1%	estimated ± 7.7 pp, low confidence
148	DeepSeek V4 Flash 0731	0.0%	estimated ± 5.1 pp, low confidence
149	Llama 4 Scout	0.0%	estimated ± 7.7 pp, low confidence
150	GPT-5.4 mini	0.0%	estimated ± 5.1 pp, low confidence
151	Gemini 1.5 Pro	0.0%	estimated ± 7.7 pp, low confidence
152	Solar Pro 3	0.0%	estimated ± 7.7 pp, low confidence
153	Qwen3-Omni-30B-A3B-Thinking	0.0%	estimated ± 7.7 pp, low confidence
154	Inkling	0.0%	estimated ± 5.1 pp, low confidence
155	Ultravox v0.6 Llama 3.3 70B	0.0%	estimated ± 7.7 pp, low confidence
156	Mistral Large 2	0.0%	estimated ± 7.7 pp, low confidence
157	Nemotron Ultra 253B	0.0%	estimated ± 7.7 pp, low confidence
158	Qwen3.6-27B	0.0%	estimated ± 5.1 pp, low confidence
159	Llama 3.1 405B	0.0%	estimated ± 7.7 pp, low confidence
160	LFM2.5-8B-A1B	0.0%	estimated ± 7.7 pp, low confidence
161	GPT-4 Turbo	0.0%	estimated ± 7.7 pp, low confidence
162	Kimi K2.5	0.0%	estimated ± 5.1 pp, low confidence
163	Kimi K2.5 (Reasoning)	0.0%	estimated ± 5.1 pp, low confidence
164	Solar Pro 2	0.0%	estimated ± 7.7 pp, low confidence
165	Nova Pro	0.0%	estimated ± 7.7 pp, low confidence
166	Qwen2.5 Coder 32B Instruct	0.0%	estimated ± 7.7 pp, low confidence
167	GPT-4o mini	0.0%	estimated ± 7.7 pp, low confidence
168	Sarvam 30B	0.0%	estimated ± 7.7 pp, low confidence
169	Celeris-1	0.0%	estimated ± 7.7 pp, low confidence
170	Exaone 4.0 32B	0.0%	estimated ± 7.7 pp, low confidence
171	Hy3 Preview	0.0%	estimated ± 5.1 pp, low confidence
172	Qwen3-Omni-30B-A3B-Instruct	0.0%	estimated ± 7.7 pp, low confidence
173	Phi-4	0.0%	estimated ± 7.7 pp, low confidence
174	Phi-4 Multimodal Instruct	0.0%	estimated ± 7.7 pp, low confidence
175	Claude Opus 4.5	0.0%	estimated ± 5.1 pp, low confidence
176	Nemotron 3 Ultra	0.0%	estimated ± 5.1 pp, low confidence
177	Claude 3 Haiku	0.0%	estimated ± 7.7 pp, low confidence
178	Gemini 1.0 Pro	0.0%	estimated ± 7.7 pp, low confidence
179	Exaone 4.0 1.2B	0.0%	estimated ± 7.7 pp, low confidence
180	Granite-4.0-H-1B	0.0%	estimated ± 7.7 pp, low confidence
181	Qwen3.5-122B-A10B	0.0%	estimated ± 5.1 pp, low confidence
182	Gemma 3 27B	0.0%	estimated ± 7.7 pp, low confidence
183	Granite-4.0-350M	0.0%	estimated ± 7.7 pp, low confidence
184	Granite-4.0-H-350M	0.0%	estimated ± 7.7 pp, low confidence
185	LFM2.5-VL-1.6B-Extract	0.0%	estimated ± 7.7 pp, low confidence
186	Ornith-1.5-9B	0.0%	estimated ± 5.1 pp, low confidence
187	GLM-5	0.0%	estimated ± 5.1 pp, low confidence
188	Qwen3.6-35B-A3B	0.0%	estimated ± 5.1 pp, low confidence
189	GLM-4.7	0.0%	estimated ± 5.1 pp, low confidence
190	Ternary Bonsai 2 27B	0.0%	estimated ± 5.1 pp, low confidence
191	Qwen3.5-27B	0.0%	estimated ± 5.1 pp, low confidence
192	Ling 3.0 Flash	0.0%	estimated ± 5.1 pp, low confidence
193	Claude 3.5 Sonnet	0.0%	estimated ± 5.1 pp, low confidence
194	Claude Sonnet 4.5	0.0%	estimated ± 5.1 pp, low confidence
195	DeepSeek V3	0.0%	estimated ± 5.1 pp, low confidence
196	Gemini 2.5 Pro	0.0%	estimated ± 5.1 pp, low confidence
197	Gemma 4 12B	0.0%	estimated ± 5.1 pp, low confidence
198	Gemma 4 31B	0.0%	estimated ± 5.1 pp, low confidence
199	Gemma 4 E2B	0.0%	estimated ± 5.1 pp, low confidence
200	Gemma 4 E4B	0.0%	estimated ± 5.1 pp, low confidence
201	GPT-4.1	0.0%	estimated ± 5.1 pp, low confidence
202	GPT-4.1 mini	0.0%	estimated ± 5.1 pp, low confidence
203	GPT-4.1 nano	0.0%	estimated ± 5.1 pp, low confidence
204	GPT-5.4 nano	0.0%	estimated ± 5.1 pp, low confidence
205	Granite 4.2 30B	0.0%	estimated ± 5.1 pp, low confidence
206	Granite 4.2 3B	0.0%	estimated ± 5.1 pp, low confidence
207	Granite 4.2 8B	0.0%	estimated ± 5.1 pp, low confidence
208	LFM2.5-230M	0.0%	estimated ± 5.1 pp, low confidence
209	LFM2.5-VL-450M	0.0%	estimated ± 5.1 pp, low confidence
210	Ling 2.6 Flash	0.0%	estimated ± 5.1 pp, low confidence
211	Ling 3.0 Flash FP8	0.0%	estimated ± 5.1 pp, low confidence
212	MAI-Thinking-1	0.0%	estimated ± 5.1 pp, low confidence
213	Mellum2-12B-A2.5B-Instruct	0.0%	estimated ± 5.1 pp, low confidence
214	Mellum2-12B-A2.5B-Thinking	0.0%	estimated ± 5.1 pp, low confidence
215	MiMo-V2-Flash	0.0%	estimated ± 5.1 pp, low confidence
216	Nemotron 3.5 Lightning 30B A3B NVFP4	0.0%	estimated ± 5.1 pp, low confidence
217	Nemotron 3 Nano Omni 30B A3B	0.0%	estimated ± 5.1 pp, low confidence
218	o1	0.0%	estimated ± 5.1 pp, low confidence
219	o1-pro	0.0%	estimated ± 5.1 pp, low confidence
220	o3-mini	0.0%	estimated ± 5.1 pp, low confidence
221	Qwen3 235B 2507	0.0%	estimated ± 5.1 pp, low confidence
222	Qwen3.5-35B-A3B	0.0%	estimated ± 5.1 pp, low confidence
223	Soofi S 30B-A3B	0.0%	estimated ± 5.1 pp, low confidence
224	ZAYA1-74B-Preview	0.0%	estimated ± 5.1 pp, low confidence
225	ZAYA1-8B	0.0%	estimated ± 5.1 pp, low confidence

HealthBench Hard leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General