benchgap
Knowledge & reasoning

BioMysteryBench (human-difficult) leaderboard

As of 2026-11-21, the highest measured score on BioMysteryBench (human-difficult) is 56.5% by Gemini 3.8 Flash. 231 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Gemini 3.8 Flash56.5%measured
2GPT-6.2 Astra51.4%estimated ± 7.8 pp, low confidence
3Claude Fable 5.250.2%estimated ± 6.8 pp, low confidence
4Claude Opus 5.550.0%measured
5Muse Wonder Full50.0%estimated ± 6.8 pp, low confidence
6GPT-5.6 Sol49.9%estimated ± 6.8 pp, low confidence
7GPT-6.2 Sol49.9%estimated ± 7.8 pp, low confidence
8GPT-6.1 Sol49.9%estimated ± 6.8 pp, low confidence
9GPT-6 Astra49.9%estimated ± 6.8 pp, low confidence
10GPT-6 Sol49.8%estimated ± 6.8 pp, low confidence
11GPT-5.5 Pro49.8%estimated ± 6.8 pp, low confidence
12GPT-5.4 Pro49.7%estimated ± 6.8 pp, low confidence
13GPT-5.6 Terra49.7%estimated ± 6.8 pp, low confidence
14Claude Fable 5.149.7%estimated ± 6.8 pp, low confidence
15Gemini 4 Neon49.6%estimated ± 7.8 pp, low confidence
16Claude Fable 549.5%estimated ± 6.8 pp, low confidence
17Muse Wonder49.5%estimated ± 6.8 pp, low confidence
18GPT-6.1 Terra49.5%estimated ± 6.8 pp, low confidence
19Claude Opus 549.4%measured
20Gemini 4 Argon49.3%estimated ± 6.8 pp, low confidence
21GPT-5.549.3%estimated ± 6.8 pp, low confidence
22MiMo-V2.6-Pro49.3%estimated ± 6.8 pp, low confidence
23Gemini 4 Carbon49.0%estimated ± 7.8 pp, low confidence
24Muse Spark 1.349.0%estimated ± 6.8 pp, low confidence
25Kimi K3.1 Pro49.0%estimated ± 6.8 pp, low confidence
26GPT-5.448.8%estimated ± 6.8 pp, low confidence
27Kimi K348.8%estimated ± 6.8 pp, low confidence
28MiniMax M3.1 Flash48.6%estimated ± 6.8 pp, low confidence
29Claude Opus 4.848.3%estimated ± 6.8 pp, low confidence
30GLM-5.248.3%estimated ± 6.8 pp, low confidence
31Step 5 Preview48.3%estimated ± 6.8 pp, low confidence
32GPT-5.6 Luna48.2%estimated ± 6.8 pp, low confidence
33Sage-A1.548.1%estimated ± 6.8 pp, low confidence
34GPT-6 Luna47.9%estimated ± 6.8 pp, low confidence
35GLM-5.347.9%estimated ± 6.8 pp, low confidence
36Claude Haiku 5.547.8%estimated ± 6.8 pp, low confidence
37DeepSeek V4 Pro47.6%estimated ± 6.8 pp, low confidence
38Ling 3.1 Flash47.6%estimated ± 6.8 pp, low confidence
39Gemini 3.1 Pro47.5%estimated ± 6.8 pp, low confidence
40Grok 4.747.5%estimated ± 6.8 pp, low confidence
41Muse Spark 1.247.5%estimated ± 6.8 pp, low confidence
42Qwen3.8 Max Preview47.5%estimated ± 6.8 pp, low confidence
43Beam 1.0 501B-A23B FP847.4%estimated ± 6.8 pp, low confidence
44Grok 4.647.3%estimated ± 6.8 pp, low confidence
45Claude Sonnet 547.2%estimated ± 6.8 pp, low confidence
46GPT-5.3 Codex47.2%estimated ± 6.8 pp, low confidence
47Hy4 preview47.2%estimated ± 6.8 pp, low confidence
48DeepSeek V4 Flash 073147.1%estimated ± 6.8 pp, low confidence
49Beam 1.0 501B-A23B47.0%estimated ± 6.8 pp, low confidence
50Gemini 4 Hydrogen46.8%estimated ± 7.8 pp, low confidence
51Grok 4.546.7%estimated ± 6.8 pp, low confidence
52Beam 1.0 501B-A23B NVFP446.7%estimated ± 6.8 pp, low confidence
53Muse Spark 1.146.6%estimated ± 6.8 pp, low confidence
54Muse Spark Open46.5%estimated ± 6.8 pp, low confidence
55DeepSeek V4.1 Flash 101746.3%estimated ± 6.8 pp, low confidence
56Qwen3.7 Max45.9%estimated ± 6.8 pp, low confidence
57Gemini 3.5 Flash45.7%estimated ± 6.8 pp, low confidence
58Claude Opus 4.6 (Adaptive)45.5%estimated ± 6.8 pp, low confidence
59Claude Opus 4.745.1%estimated ± 6.8 pp, low confidence
60MiMo-V2.6-Flash45.1%estimated ± 6.8 pp, low confidence
61GPT-5.244.9%estimated ± 6.8 pp, low confidence
62Muse Spark44.7%estimated ± 6.8 pp, low confidence
63Claude Sonnet 5.544.7%measured
64Qwen3.8-Flash-Next44.6%estimated ± 6.8 pp, low confidence
65Qwen4 Max44.4%estimated ± 7.8 pp, low confidence
66Nemotron 3.5 Super VL 120B A12B44.4%estimated ± 6.8 pp, low confidence
67Gemini 3.6 Flash44.3%estimated ± 6.8 pp, low confidence
68Nemotron 3.5 Super VL 120B A12B NVFP444.3%estimated ± 6.8 pp, low confidence
69GPT-5.4 mini43.8%estimated ± 6.8 pp, low confidence
70Kimi K2.7 Code43.8%estimated ± 6.8 pp, low confidence
71Mistral Large 443.8%estimated ± 6.8 pp, low confidence
72Gemini 3.7 Flash43.5%measured
73Quasar 438B43.3%estimated ± 6.8 pp, low confidence
74GPT-5.4 nano43.2%estimated ± 6.8 pp, low confidence
75o3-pro43.1%estimated ± 8.3 pp, low confidence
76Gemini 3 Pro43.0%estimated ± 6.8 pp, low confidence
77Qwen3.7 Plus43.0%estimated ± 6.8 pp, low confidence
78A.X K242.9%estimated ± 6.8 pp, low confidence
79GPT-5.2-Codex42.7%estimated ± 6.8 pp, low confidence
80Inkling-Small42.3%estimated ± 6.8 pp, low confidence
81Grok 4.341.9%estimated ± 6.8 pp, low confidence
82Kimi K2.641.9%estimated ± 6.8 pp, low confidence
83Claude 4.1 Opus41.6%estimated ± 8.3 pp, low confidence
84Claude 4 Opus40.5%estimated ± 8.3 pp, low confidence
85Inkling 2 Small40.0%estimated ± 7.8 pp, low confidence
86Solar Pro 439.7%estimated ± 6.8 pp, low confidence
87GPT-538.6%estimated ± 6.8 pp, low confidence
88GPT-5.1-Codex38.6%estimated ± 6.8 pp, low confidence
89GPT-5.1-Codex-Max38.6%estimated ± 6.8 pp, low confidence
90Inkling38.0%estimated ± 6.8 pp, low confidence
91Qwen3.8-27B38.0%estimated ± 6.8 pp, low confidence
92o1-pro37.3%estimated ± 8.3 pp, low confidence
93GPT-5.137.0%estimated ± 6.8 pp, low confidence
94Hy337.0%estimated ± 6.8 pp, low confidence
95Hy3 Preview37.0%estimated ± 6.8 pp, low confidence
96Apodex 1.1 Mini36.2%estimated ± 6.8 pp, low confidence
97Claude Opus 4.5 Thinking36.2%estimated ± 6.8 pp, low confidence
98GLM-5.136.2%estimated ± 6.8 pp, low confidence
99o1-preview36.2%estimated ± 8.3 pp, low confidence
100Nemotron 3.5 Bolt 74B-A9B35.5%estimated ± 6.8 pp, low confidence
101Apodex 1.135.3%measured
102MiMo-V2.5-Pro34.6%estimated ± 6.8 pp, low confidence
103o1-mini34.4%estimated ± 8.3 pp, low confidence
104GPT-4.534.1%estimated ± 8.3 pp, low confidence
105MiniMax M333.6%estimated ± 6.8 pp, low confidence
106Qwen 3.6 Max (preview)33.6%estimated ± 6.8 pp, low confidence
107Claude 3 Opus32.9%estimated ± 8.3 pp, low confidence
108DeepSeek R1 Distill Qwen 32B32.4%estimated ± 8.3 pp, low confidence
109Nemotron 3.5 Bolt 74B-A9B BF1632.3%estimated ± 7.8 pp, low confidence
110Muse Ember31.7%estimated ± 7.8 pp, low confidence
111Gemini 1.5 Pro31.6%estimated ± 8.3 pp, low confidence
112Kimi K2.531.4%estimated ± 6.8 pp, low confidence
113Nemotron 3 Ultra31.4%estimated ± 6.8 pp, low confidence
114Inkling 2 Mini31.2%estimated ± 7.8 pp, low confidence
115Grok 4.1 Fast (Reasoning)30.5%estimated ± 6.8 pp, low confidence
116Grok 4 Fast (Reasoning)30.5%estimated ± 6.8 pp, low confidence
117Qwen3.6 Plus30.5%estimated ± 6.8 pp, low confidence
118Gemini 1.5 Flash30.1%estimated ± 8.3 pp, low confidence
119Claude Opus 4.630.1%estimated ± 6.8 pp, low confidence
120GPT-4 Turbo30.1%estimated ± 8.3 pp, low confidence
121Qwen2.5 Coder 32B Instruct29.5%estimated ± 8.3 pp, low confidence
122GPT-4o mini29.3%estimated ± 8.3 pp, low confidence
123DeepSeek-V2.529.2%estimated ± 8.3 pp, low confidence
124Gemini 2.5 Pro29.1%estimated ± 6.8 pp, low confidence
125Muse Glimmer 30B29.1%estimated ± 6.8 pp, low confidence
126Step 3.7 Flash27.5%estimated ± 6.8 pp, low confidence
127Gemini 1.0 Pro26.3%estimated ± 8.3 pp, low confidence
128GLM-525.6%estimated ± 6.8 pp, low confidence
129Grok 425.6%estimated ± 6.8 pp, low confidence
130Ling 3.0 Flash VL25.6%estimated ± 6.8 pp, low confidence
131GLM-4.723.5%estimated ± 6.8 pp, low confidence
132Ling 3.0 Flash23.5%estimated ± 6.8 pp, low confidence
133Qwen3.5 397B23.5%estimated ± 6.8 pp, low confidence
134DeepSeek-R120.9%estimated ± 6.8 pp, low confidence
135Gemini 2.5 Flash20.9%estimated ± 6.8 pp, low confidence
136Gemini 3 Flash20.9%estimated ± 6.8 pp, low confidence
137Gemma 4 31B20.9%estimated ± 6.8 pp, low confidence
138GPT-OSS 20B20.9%estimated ± 6.8 pp, low confidence
139Claude 4 Opus Thinking19.4%estimated ± 8.1 pp, low confidence
140Claude 4 Sonnet18.0%estimated ± 6.8 pp, low confidence
141GPT-OSS 120B18.0%estimated ± 6.8 pp, low confidence
142K-Exaone18.0%estimated ± 6.8 pp, low confidence
143MiMo-V2-Omni18.0%estimated ± 6.8 pp, low confidence
144o318.0%estimated ± 6.8 pp, low confidence
145Qwen3.6-27B18.0%estimated ± 6.8 pp, low confidence
146Claude 3.7 Sonnet Thinking15.7%estimated ± 6.8 pp, low confidence
147Claude Sonnet 4.615.7%estimated ± 6.8 pp, low confidence
148DeepSeek V3.215.7%estimated ± 6.8 pp, low confidence
149K-EXAONE 2.015.7%estimated ± 6.8 pp, low confidence
150Qwen3.5-27B15.7%estimated ± 6.8 pp, low confidence
151Qwen3.5-35B-A3B15.7%estimated ± 6.8 pp, low confidence
152Trinity-Large-Preview15.7%estimated ± 6.8 pp, low confidence
153Trinity-Large-Thinking15.7%estimated ± 6.8 pp, low confidence
154LFM3-1.7B15.1%estimated ± 7.8 pp, low confidence
155o3-mini13.5%estimated ± 8.1 pp, low confidence
156QwQ-32B12.8%estimated ± 8.1 pp, low confidence
157Gemma 4 E4B11.6%estimated ± 6.8 pp, low confidence
158GLM-5V-Turbo11.6%estimated ± 6.8 pp, low confidence
159MiniMax M2.711.6%estimated ± 6.8 pp, low confidence
160o4-mini11.6%estimated ± 6.8 pp, low confidence
161Qwen3.5-122B-A10B11.6%estimated ± 6.8 pp, low confidence
162Claude Opus 4.56.5%estimated ± 6.8 pp, low confidence
163Command A+6.5%estimated ± 6.8 pp, low confidence
164GLM-5-Turbo6.5%estimated ± 6.8 pp, low confidence
165Granite 4.2 30B6.5%estimated ± 6.8 pp, low confidence
166Granite 4.2 8B6.5%estimated ± 6.8 pp, low confidence
167Magistral Medium6.5%estimated ± 6.8 pp, low confidence
168MiMo-V2-Pro6.5%estimated ± 6.8 pp, low confidence
169MiniCPM5-2B6.5%estimated ± 6.8 pp, low confidence
170Mistral Small 46.5%estimated ± 6.8 pp, low confidence
171North Mini Code6.5%estimated ± 6.8 pp, low confidence
172o16.5%estimated ± 6.8 pp, low confidence
173Qwen3.6-35B-A3B6.5%estimated ± 6.8 pp, low confidence
174Sarvam 30B6.5%estimated ± 6.8 pp, low confidence
175Celeris-10.0%estimated ± 6.8 pp, low confidence
176Claude 3.5 Haiku0.0%estimated ± 6.8 pp, low confidence
177Claude 3.7 Sonnet0.0%estimated ± 6.8 pp, low confidence
178Claude 3 Haiku0.0%estimated ± 6.8 pp, low confidence
179DeepSeek V30.0%estimated ± 6.8 pp, low confidence
180DeepSeek V3 03240.0%estimated ± 6.8 pp, low confidence
181DeepSeek V3.10.0%estimated ± 6.8 pp, low confidence
182Devstral Small0.0%estimated ± 6.8 pp, low confidence
183Exaone 4.0 1.2B0.0%estimated ± 6.8 pp, low confidence
184Exaone 4.0 32B0.0%estimated ± 6.8 pp, low confidence
185Gemini 2.0 Flash0.0%estimated ± 6.8 pp, low confidence
186Gemini 2.5 Flash-Lite0.0%estimated ± 6.8 pp, low confidence
187Gemini 3.5 Flash-Lite0.0%estimated ± 6.8 pp, low confidence
188Gemma 3 27B0.0%estimated ± 6.8 pp, low confidence
189Gemma 4 12B0.0%estimated ± 6.8 pp, low confidence
190Gemma 4 26B A4B0.0%estimated ± 6.8 pp, low confidence
191Gemma 4 E2B0.0%estimated ± 6.8 pp, low confidence
192GLM-4.5-Air0.0%estimated ± 6.8 pp, low confidence
193GLM-4.60.0%estimated ± 6.8 pp, low confidence
194GPT-4.10.0%estimated ± 6.8 pp, low confidence
195GPT-4.1 mini0.0%estimated ± 6.8 pp, low confidence
196GPT-4.1 nano0.0%estimated ± 6.8 pp, low confidence
197GPT-4o0.0%estimated ± 6.8 pp, low confidence
198Granite-4.0-350M0.0%estimated ± 6.8 pp, low confidence
199Granite-4.0-H-1B0.0%estimated ± 6.8 pp, low confidence
200Granite-4.0-H-350M0.0%estimated ± 6.8 pp, low confidence
201Granite 4.2 3B0.0%estimated ± 6.8 pp, low confidence
202Grok 4.1 Fast0.0%estimated ± 6.8 pp, low confidence
203Grok Code Fast 10.0%estimated ± 6.8 pp, low confidence
204Kimi K20.0%estimated ± 6.8 pp, low confidence
205LFM2.5-2.6B0.0%estimated ± 6.8 pp, low confidence
206LFM2.5-8B-A1B0.0%estimated ± 6.8 pp, low confidence
207LFM2.5-VL-1.6B-Extract0.0%estimated ± 6.8 pp, low confidence
208Ling 2.6 Flash0.0%estimated ± 6.8 pp, low confidence
209Ling 3.0 Tiny0.0%estimated ± 6.8 pp, low confidence
210Llama 3.1 405B0.0%estimated ± 6.8 pp, low confidence
211Llama 3.1 70B0.0%estimated ± 6.8 pp, low confidence
212Llama 3.3 70B0.0%estimated ± 6.8 pp, low confidence
213Llama 4 Maverick0.0%estimated ± 6.8 pp, low confidence
214Llama 4 Scout0.0%estimated ± 6.8 pp, low confidence
215Magistral Small0.0%estimated ± 6.8 pp, low confidence
216Mercury 2.50.0%estimated ± 6.8 pp, low confidence
217MiMo-V2-Flash0.0%estimated ± 6.8 pp, low confidence
218Mistral Large 20.0%estimated ± 6.8 pp, low confidence
219Mistral Large 30.0%estimated ± 6.8 pp, low confidence
220Mistral Medium 30.0%estimated ± 6.8 pp, low confidence
221Mistral Medium 3.5 128B0.0%estimated ± 6.8 pp, low confidence
222Mistral Small 3.10.0%estimated ± 6.8 pp, low confidence
223Nemotron 3.5 Lightning 30B A3B NVFP40.0%estimated ± 6.8 pp, low confidence
224Nemotron 3 Nano 30B0.0%estimated ± 6.8 pp, low confidence
225Nemotron 3 Nano Omni 30B A3B0.0%estimated ± 6.8 pp, low confidence
226Nemotron Ultra 253B0.0%estimated ± 6.8 pp, low confidence
227Nova Pro0.0%estimated ± 6.8 pp, low confidence
228Phi-40.0%estimated ± 6.8 pp, low confidence
229Phi-4-reasoning0.0%estimated ± 6.8 pp, low confidence
230Qwen3.5-4B0.0%estimated ± 6.8 pp, low confidence
231Qwen3 Max0.0%estimated ± 6.8 pp, low confidence
232Qwen3-Omni-30B-A3B-Instruct0.0%estimated ± 6.8 pp, low confidence
233Qwen3-Omni-30B-A3B-Thinking0.0%estimated ± 6.8 pp, low confidence
234Sarvam 105B0.0%estimated ± 6.8 pp, low confidence
235Solar Pro 20.0%estimated ± 6.8 pp, low confidence
236Solar Pro 30.0%estimated ± 6.8 pp, low confidence
237Ultravox v0.6 Llama 3.3 70B0.0%estimated ± 6.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General