benchgap
Knowledge & reasoning

HLE leaderboard

As of 2026-11-21, the highest measured score on HLE is 74.2% by Claude Fable 5.2. 231 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.2 Astra74.5%estimated ± 2.2 pp, medium confidence
2Claude Fable 5.274.2%measured
3Gemini 4 Carbon72.6%estimated ± 2.2 pp, high confidence
4GPT-6 Astra72.3%estimated ± 2.2 pp, high confidence
5GPT-6.2 Sol70.0%estimated ± 2.2 pp, high confidence
6Kimi K3.1 Pro70.0%measured
7GPT-6.1 Sol69.4%estimated ± 2.2 pp, high confidence
8Gemini 4 Argon67.5%estimated ± 2.2 pp, high confidence
9Claude Fable 5.165.0%measured
10Claude Opus 564.7%measured
11Claude Mythos 564.5%measured
12Claude Opus 5.563.8%estimated ± 6.1 pp, low confidence
13Claude Sonnet 5.562.6%estimated ± 6.5 pp, medium confidence
14Muse Wonder Full62.4%estimated ± 8.5 pp, medium confidence
15Muse Spark 1.162.1%measured
16Claude Fable 561.4%estimated ± 6.1 pp, medium confidence
17Gemini 4 Neon61.0%estimated ± 10.0 pp, low confidence
18Hy459.1%measured
19GPT-5.4 Pro58.7%measured
20Muse Wonder58.3%estimated ± 8.5 pp, medium confidence
21Claude Opus 4.857.9%measured
22Gemini 3.7 Flash57.5%estimated ± 6.1 pp, medium confidence
23MiMo-V2.6-Pro57.5%estimated ± 8.5 pp, medium confidence
24Claude Sonnet 557.4%measured
25GPT-5.5 Pro57.2%measured
26Sakana Fugu-Ultra57.2%estimated ± 6.5 pp, medium confidence
27Gemini 3.8 Flash57.2%estimated ± 2.2 pp, high confidence
28Muse Spark 1.356.6%estimated ± 8.5 pp, medium confidence
29GPT-5.6 Sol56.4%estimated ± 2.2 pp, high confidence
30Pareto 26.956.4%estimated ± 6.5 pp, medium confidence
31Apodex 1.156.1%measured
32Kimi K356.0%measured
33GPT-6 Sol55.5%estimated ± 2.2 pp, high confidence
34Hy4 preview55.4%measured
35Hy4 flash55.2%measured
36Sakana Fugu54.9%estimated ± 6.5 pp, medium confidence
37Grok 4.654.8%estimated ± 2.2 pp, high confidence
38Grok 4.754.7%estimated ± 2.2 pp, high confidence
39Claude Opus 4.754.7%measured
40GPT-5.6 Terra54.4%estimated ± 2.2 pp, high confidence
41Claude Haiku 5.554.3%estimated ± 2.2 pp, high confidence
42Gemini 3.1 Pro54.3%estimated ± 2.2 pp, high confidence
43Grok 4.554.3%estimated ± 2.2 pp, high confidence
44GPT-5.6 Luna54.2%estimated ± 2.2 pp, high confidence
45GPT-6 Luna54.2%estimated ± 2.2 pp, medium confidence
46Grok 4.2054.2%estimated ± 2.2 pp, medium confidence
47ERNIE 5.5 Flash Preview54.2%estimated ± 6.5 pp, medium confidence
48GPT-6.1 Terra53.3%estimated ± 8.5 pp, medium confidence
49Inkling 2 Small53.2%measured
50Claude Opus 4.653.0%measured
51dots3-note Preview52.6%measured
52Gemini 4 Hydrogen52.4%estimated ± 10.0 pp, low confidence
53Muse Spark 1.252.4%estimated ± 8.5 pp, medium confidence
54GPT-5.552.2%measured
55GPT-5.452.1%measured
56Qwen4 Max51.7%measured
57Sage-A1.551.7%estimated ± 10.4 pp, low confidence
58MiniMax M3.1 Flash51.6%estimated ± 6.5 pp, medium confidence
59Muse Spark50.4%measured
60Muse Spark Open50.2%estimated ± 8.5 pp, medium confidence
61Qwen3.8 Max Preview49.1%estimated ± 8.5 pp, medium confidence
62Claude Sonnet 4.649.0%measured
63Claude Opus 4.6 (Adaptive)48.4%estimated ± 6.1 pp, medium confidence
64GPT-5.3 Codex48.2%estimated ± 8.5 pp, medium confidence
65MiMo-V2.5-Pro48.0%measured
66GLM-5.348.0%estimated ± 8.5 pp, medium confidence
67Inkling-Small47.8%measured
68Pareto 26.10 Preview47.7%estimated ± 6.8 pp, medium confidence
69Agents-A147.6%measured
70Step 5 Preview46.5%measured
71Inkling46.0%measured
72Gemini 3.6 Flash46.0%estimated ± 6.1 pp, medium confidence
73GPT-5.244.8%estimated ± 6.1 pp, medium confidence
74Beam 1.0 501B-A23B44.8%estimated ± 6.5 pp, medium confidence
75Beam 1.0 501B-A23B FP844.8%estimated ± 6.5 pp, medium confidence
76Ornith-1.5-397B44.6%measured
77Claude Opus 4.5 Thinking44.0%estimated ± 6.1 pp, medium confidence
78Gemini 3 Pro43.9%estimated ± 6.1 pp, medium confidence
79Claude Sonnet 4.5 Thinking43.8%estimated ± 6.1 pp, medium confidence
80Claude Sonnet 4.543.8%estimated ± 6.1 pp, medium confidence
81Qwen3.8 Max43.6%measured
82Beam 1.0 501B-A23B NVFP443.6%estimated ± 6.5 pp, medium confidence
83MiniMax M343.4%estimated ± 8.5 pp, medium confidence
84DeepSeek V4 Pro42.7%measured
85DeepSeek V4.1 Flash 101742.6%measured
86GPT-5.142.1%estimated ± 7.0 pp, medium confidence
87GPT-5.4 mini41.5%measured
88Qwen3.7 Max41.4%measured
89GLM-5.240.5%measured
90Gemini 3.1 Flash-Lite40.4%estimated ± 9.2 pp, low confidence
91Gemini 3.5 Flash40.2%measured
92GLM-5.3-Flash40.1%estimated ± 9.2 pp, low confidence
93Particle-140.0%estimated ± 6.8 pp, medium confidence
94Inkling 2 Mini39.5%estimated ± 6.5 pp, medium confidence
95GPT-5.2-Codex39.0%estimated ± 8.5 pp, medium confidence
96Qwen 3.6 Max (preview)38.5%estimated ± 5.1 pp, medium confidence
97MiMo-V2.6-Flash38.2%estimated ± 8.5 pp, medium confidence
98Kimi K2.7 Code38.1%estimated ± 8.5 pp, medium confidence
99GPT-5.4 nano37.7%measured
100Solar Open 237.5%estimated ± 6.5 pp, medium confidence
101Ling 3.1 Flash37.4%measured
102Nemotron 3.5 Bolt 74B-A9B BF1637.4%estimated ± 6.5 pp, medium confidence
103Apodex 1.1 Mini37.0%estimated ± 8.5 pp, medium confidence
104Interfaze Beta36.9%estimated ± 6.5 pp, medium confidence
105Qwen3.8-Omni-Flash36.5%measured
106Hy336.2%estimated ± 8.5 pp, medium confidence
107Mercury 335.9%estimated ± 6.8 pp, medium confidence
108Qwen3.8-Flash-Next35.9%measured
109Qwen3.5 Flash35.2%estimated ± 9.2 pp, low confidence
110Solar Pro 435.2%estimated ± 6.8 pp, medium confidence
111Grok 4.335.0%measured
112DeepSeek V4 Flash 073134.8%measured
113Kimi K2.634.7%measured
114Qwen3.7 Plus34.7%measured
115Nemotron 3.5 Bolt 74B-A9B33.7%estimated ± 6.5 pp, medium confidence
116OrcaCyber-Nano-1.032.9%estimated ± 6.8 pp, medium confidence
117MiMo-V2.532.8%estimated ± 9.2 pp, low confidence
118Claude 4.1 Opus32.7%estimated ± 12.0 pp, low confidence
119Seed 2.1 Pro32.7%estimated ± 5.1 pp, medium confidence
120MiMo-V2-Pro32.6%estimated ± 8.5 pp, medium confidence
121Mistral Large 431.3%estimated ± 10.0 pp, low confidence
122GLM-5.131.0%measured
123Claude Opus 4.530.8%measured
124Qwen3.8-27B30.8%measured
125GPT-530.5%estimated ± 8.5 pp, medium confidence
126GLM-530.5%measured
127Kimi K2.530.1%measured
128GLM-5-Turbo29.8%estimated ± 8.5 pp, medium confidence
129GLM-4.529.7%estimated ± 9.2 pp, low confidence
130o3-pro29.6%estimated ± 10.1 pp, low confidence
131Nemotron 3.5 Super VL 120B A12B29.2%measured
132Nemotron 3.5 Super VL 120B A12B NVFP429.0%measured
133Qwen3.6 Plus28.8%measured
134Grok 428.7%estimated ± 8.5 pp, medium confidence
135Qwen3.5 397B28.7%measured
136MiniMax M2.728.0%measured
137GPT-5.1-Codex27.8%estimated ± 8.5 pp, medium confidence
138GPT-5.1-Codex-Max27.8%estimated ± 8.5 pp, medium confidence
139A.X K227.8%measured
140Muse Ember27.1%estimated ± 10.0 pp, low confidence
141Nemotron 3 Ultra26.7%measured
142Gemma 4 31B26.5%measured
143Claude Haiku 4.525.9%estimated ± 9.2 pp, low confidence
144Ternary Bonsai 2 27B25.7%estimated ± 6.5 pp, medium confidence
145Seed 2.1 Turbo25.7%estimated ± 5.1 pp, medium confidence
146Ornith-1.5-35B-A3B25.6%measured
147Hy3 Preview25.5%measured
148Qwen3.5-122B-A10B25.1%estimated ± 5.1 pp, medium confidence
149Spark-X2.6-4B25.1%estimated ± 6.5 pp, medium confidence
150GLM-4.724.8%measured
151MiMo-V2-Omni24.8%estimated ± 8.5 pp, medium confidence
152Ling 3.0 Flash VL24.7%estimated ± 8.5 pp, medium confidence
153Muse Glimmer 30B24.7%estimated ± 8.5 pp, medium confidence
154Ripple-65B Preview24.4%estimated ± 6.8 pp, medium confidence
155Step 3.7 Flash24.3%estimated ± 8.5 pp, medium confidence
156Qwen3.6-27B24.0%measured
157o323.4%estimated ± 8.5 pp, medium confidence
158Grok 4.1 Fast (Reasoning)22.8%estimated ± 8.5 pp, medium confidence
159MAI-Thinking-122.8%estimated ± 6.5 pp, medium confidence
160Grok 4 Fast (Reasoning)22.7%estimated ± 8.5 pp, medium confidence
161Ling 3.0 Flash22.7%measured
162Gemini 3.5 Flash-Lite22.5%estimated ± 8.5 pp, medium confidence
163Ling 3.0 Flash FP822.5%estimated ± 6.5 pp, medium confidence
164Quasar 438B22.5%estimated ± 8.5 pp, medium confidence
165Qwen3.5-27B22.1%estimated ± 5.1 pp, medium confidence
166MiMo-V2-Flash22.0%estimated ± 6.5 pp, medium confidence
167GLM-5V-Turbo21.5%estimated ± 8.5 pp, medium confidence
168Qwen3.6-35B-A3B21.4%measured
169Qwen3 235B-A22B (Reasoning)21.2%estimated ± 2.3 pp, low confidence
170Gemini 3 Flash20.4%estimated ± 8.5 pp, medium confidence
171Ornith-1.5-9B20.2%measured
172K-Exaone19.9%estimated ± 8.5 pp, medium confidence
173Mistral Medium 3.5 128B19.8%estimated ± 8.5 pp, medium confidence
174o1-preview19.8%estimated ± 10.1 pp, low confidence
175Command A+19.1%estimated ± 8.5 pp, medium confidence
176Qwen3 Max19.0%estimated ± 8.5 pp, medium confidence
177Gemini 2.5 Pro18.8%measured
178DeepSeek V3.218.8%estimated ± 8.5 pp, medium confidence
179North Mini Code18.7%estimated ± 8.5 pp, medium confidence
180QwQ-32B18.6%estimated ± 2.3 pp, low confidence
181Solar Pro 318.4%estimated ± 8.5 pp, medium confidence
182K-EXAONE 2.018.3%measured
183Mistral Small 418.3%estimated ± 8.5 pp, medium confidence
184Qwen3.5-4B18.3%estimated ± 8.5 pp, medium confidence
185Nemotron 3 Super 100B18.3%measured
186Qwen3.5-35B-A3B18.1%estimated ± 5.1 pp, medium confidence
187Ling 3.0 Tiny18.1%estimated ± 8.5 pp, medium confidence
188DeepSeek-R1-052817.7%measured
189Grok Code Fast 117.7%estimated ± 8.5 pp, medium confidence
190Qwen3-Omni-30B-A3B-Thinking17.5%estimated ± 8.5 pp, medium confidence
191GLM-4.5-Air17.4%estimated ± 8.5 pp, medium confidence
192LFM2.5-8B-A1B17.4%estimated ± 8.5 pp, medium confidence
193Celeris-117.3%estimated ± 8.5 pp, medium confidence
194Grok 3 Mini17.3%estimated ± 6.5 pp, medium confidence
195Granite-4.0-H-350M17.2%estimated ± 8.5 pp, medium confidence
196Gemma 4 26B A4B17.2%measured
197LFM2.5-2.6B17.2%estimated ± 8.5 pp, medium confidence
198Exaone 4.0 1.2B17.0%estimated ± 8.5 pp, medium confidence
199GLM-4.617.0%estimated ± 8.5 pp, medium confidence
200Granite-4.0-350M17.0%estimated ± 8.5 pp, medium confidence
201Grok 4.1 Fast16.9%estimated ± 8.5 pp, medium confidence
202LFM2.5-VL-1.6B-Extract16.9%estimated ± 8.5 pp, medium confidence
203Qwen3 235B 250716.9%estimated ± 5.1 pp, medium confidence
204Exaone 4.0 32B16.9%estimated ± 8.5 pp, medium confidence
205Granite-4.0-H-1B16.9%estimated ± 8.5 pp, medium confidence
206Phi-4 Multimodal Instruct16.9%estimated ± 8.5 pp, medium confidence
207Llama 4 Maverick16.8%estimated ± 8.5 pp, medium confidence
208DeepSeek V3 032416.8%estimated ± 8.5 pp, medium confidence
209Gemini 2.5 Flash16.8%estimated ± 8.5 pp, medium confidence
210DeepSeek R1 Distill Qwen 32B16.8%estimated ± 8.5 pp, medium confidence
211Gemini 1.5 Pro16.8%estimated ± 8.5 pp, medium confidence
212Qwen3-Omni-30B-A3B-Instruct16.8%estimated ± 8.5 pp, medium confidence
213Gemma 3 27B16.7%estimated ± 8.5 pp, low confidence
214Claude 4 Sonnet16.7%estimated ± 8.5 pp, low confidence
215Llama 3.1 8B16.7%estimated ± 8.5 pp, low confidence
216Gemini 1.0 Pro16.7%estimated ± 8.5 pp, low confidence
217GPT-4o mini16.7%estimated ± 8.5 pp, low confidence
218Mistral Large 316.7%estimated ± 8.5 pp, low confidence
219Claude 3 Haiku16.7%estimated ± 8.5 pp, low confidence
220Mistral Medium 316.7%estimated ± 8.5 pp, low confidence
221Llama 3.1 405B16.6%estimated ± 8.5 pp, low confidence
222Devstral Small16.6%estimated ± 8.5 pp, low confidence
223Llama 4 Scout16.6%estimated ± 8.5 pp, low confidence
224Solar Pro 216.6%estimated ± 8.5 pp, low confidence
225Ultravox v0.6 Llama 3.3 70B16.5%estimated ± 8.5 pp, low confidence
226Qwen2.5 Coder 32B Instruct16.5%estimated ± 8.5 pp, low confidence
227Laguna XS.216.5%estimated ± 9.2 pp, low confidence
228Claude 4 Opus Thinking16.5%estimated ± 6.5 pp, medium confidence
229Mistral Large 216.5%estimated ± 8.5 pp, low confidence
230Nova Pro16.5%estimated ± 8.5 pp, low confidence
231GPT-4 Turbo16.4%estimated ± 8.5 pp, low confidence
232Claude 3 Opus16.4%estimated ± 8.5 pp, low confidence
233GPT-4o16.3%estimated ± 8.5 pp, low confidence
234Laguna M.116.3%estimated ± 9.2 pp, low confidence
235Qwen3 32B (Reasoning)16.1%estimated ± 2.3 pp, low confidence
236o1-pro15.9%estimated ± 6.5 pp, medium confidence
237Nemotron 3.5 Lightning 30B A3B NVFP415.5%estimated ± 6.5 pp, medium confidence
238Claude 3.7 Sonnet Thinking15.0%estimated ± 6.5 pp, medium confidence
239GPT-OSS 120B14.9%measured
240o4-mini14.5%estimated ± 6.5 pp, medium confidence
241o3-mini14.1%estimated ± 6.5 pp, medium confidence
242Mercury 2.514.0%estimated ± 6.8 pp, medium confidence
243Sarvam 105B13.7%estimated ± 6.8 pp, medium confidence
244Qwen3 235B-A22B13.4%estimated ± 2.3 pp, low confidence
245Nemotron Ultra 253B13.1%estimated ± 6.5 pp, medium confidence
246o112.8%estimated ± 6.5 pp, medium confidence
247Grok 3 [Beta]12.6%estimated ± 6.5 pp, medium confidence
248GLM-4.7-Flash12.4%estimated ± 6.5 pp, medium confidence
249Claude 4 Opus12.2%estimated ± 6.5 pp, medium confidence
250DeepSeek V3.112.2%estimated ± 6.5 pp, medium confidence
251Trinity-Large-Thinking11.5%estimated ± 6.8 pp, medium confidence
252GPT-OSS 20B10.9%measured
253Nemotron 3 Nano 30B10.6%measured
254Nemotron 3 Nano Omni 30B A3B10.3%estimated ± 6.5 pp, medium confidence
255DeepSeek-R19.9%estimated ± 6.5 pp, medium confidence
256ZAYA1-8B9.6%estimated ± 6.5 pp, medium confidence
257Magistral Medium9.5%estimated ± 6.5 pp, medium confidence
258MiniCPM5-2B8.9%measured
259GPT-4.58.8%estimated ± 6.5 pp, medium confidence
260Phi-4-reasoning-plus8.5%estimated ± 6.5 pp, medium confidence
261LongCat-Flash-Lite-Sparse8.2%estimated ± 6.8 pp, medium confidence
262Magistral Small8.1%estimated ± 6.5 pp, medium confidence
263Claude 3.7 Sonnet8.0%estimated ± 6.5 pp, medium confidence
264Gemma 4 12B8.0%estimated ± 6.5 pp, low confidence
265Sarvam 30B7.5%estimated ± 6.8 pp, medium confidence
266LFM2.5-24B-A2B7.4%measured
267Granite 4.2 30B7.3%estimated ± 6.5 pp, medium confidence
268GPT-4.17.2%estimated ± 6.5 pp, medium confidence
269LFM3-1.7B7.2%measured
270Mellum2.1-12B-A2.5B-Thinking7.2%estimated ± 6.8 pp, medium confidence
271Phi-4-reasoning7.0%estimated ± 6.5 pp, medium confidence
272Trinity-Large-Preview7.0%estimated ± 6.8 pp, low confidence
273DeepSeek R1 Distill Llama 70B6.8%estimated ± 6.5 pp, medium confidence
274Gemini 2.0 Flash6.8%estimated ± 6.5 pp, medium confidence
275Gemini 2.5 Flash-Lite6.5%estimated ± 6.5 pp, medium confidence
276GPT-4.1 mini6.4%estimated ± 6.5 pp, medium confidence
277Granite 4.2 8B6.3%estimated ± 6.5 pp, medium confidence
278LLaDA2.2-mini6.3%estimated ± 6.8 pp, low confidence
279MiniCPM5-1B5.8%estimated ± 5.1 pp, low confidence
280DeepSeek-R1-0528-Qwen3-8B5.2%estimated ± 6.5 pp, low confidence
281DeepSeek-V2.55.2%estimated ± 10.1 pp, low confidence
282Qwen2.5-Max4.9%estimated ± 6.5 pp, low confidence
283o1-mini4.9%estimated ± 6.5 pp, low confidence
284Kimi K24.7%measured
285Claude 3.5 Sonnet4.7%estimated ± 6.5 pp, low confidence
286DeepSeek R1 Distill Qwen 14B4.6%estimated ± 6.5 pp, low confidence
287DeepSeek V34.6%estimated ± 6.5 pp, low confidence
288Ling 2.6 Flash4.5%estimated ± 6.5 pp, low confidence
289Gemma 4 E4B4.4%estimated ± 6.5 pp, low confidence
290Mellum2-12B-A2.5B-Thinking4.1%estimated ± 6.5 pp, low confidence
291ZAYA1-74B-Preview4.0%estimated ± 6.5 pp, low confidence
292Phi-43.7%estimated ± 6.5 pp, low confidence
293Grok 23.7%estimated ± 6.5 pp, low confidence
294Granite 4.2 3B3.4%estimated ± 6.5 pp, low confidence
295Qwen3 32B3.3%estimated ± 6.5 pp, low confidence
296MiniMax-Text-013.3%estimated ± 6.5 pp, low confidence
297Grok 2 mini2.5%estimated ± 6.5 pp, low confidence
298Gemini 2.0 Flash-Lite2.4%estimated ± 6.5 pp, low confidence
299Llama 3.3 70B2.4%estimated ± 6.5 pp, low confidence
300GPT-4.1 nano2.3%estimated ± 6.5 pp, low confidence
301Gemini 1.5 Flash2.3%estimated ± 6.5 pp, low confidence
302Llama 3.1 70B1.6%estimated ± 6.5 pp, low confidence
303Qwen2.5-VL-32B1.5%estimated ± 6.5 pp, low confidence
304Mistral Small 3.11.5%estimated ± 6.5 pp, low confidence
305Gemma 4 E2B1.0%estimated ± 6.5 pp, low confidence
306Soofi S 30B-A3B1.0%estimated ± 6.5 pp, low confidence
307Claude 3.5 Haiku0.7%estimated ± 6.5 pp, low confidence
308Mellum2-12B-A2.5B-Instruct0.6%estimated ± 6.5 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General