benchgap
Knowledge & reasoning

HLE w/o tools leaderboard

As of 2026-11-21, the highest measured score on HLE w/o tools is 73.2% by Claude Fable 5.2. 247 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.273.2%measured
2Gemini 4 Neon65.0%estimated ± 4.9 pp, high confidence
3Claude Opus 5.564.4%measured
4Gemini 4 Carbon63.9%estimated ± 4.9 pp, high confidence
5Claude Fable 5.160.9%measured
6GPT-6.1 Sol59.6%estimated ± 4.9 pp, high confidence
7Gemini 4 Hydrogen59.0%estimated ± 4.9 pp, high confidence
8Claude Mythos 559.0%measured
9Ling 3.1 Flash57.0%estimated ± 4.9 pp, medium confidence
10GPT-6.1 Terra57.0%estimated ± 4.9 pp, high confidence
11Claude Sonnet 5.556.9%measured
12Claude Opus 556.3%measured
13Gemini 4 Argon56.2%estimated ± 4.9 pp, high confidence
14Muse Wonder Full56.0%estimated ± 4.9 pp, high confidence
15Claude Fable 554.6%estimated ± 4.9 pp, high confidence
16Muse Spark 1.354.4%estimated ± 4.9 pp, high confidence
17GPT-6 Sol54.0%estimated ± 4.9 pp, high confidence
18MiMo-V2.6-Pro53.0%estimated ± 4.9 pp, high confidence
19Muse Spark 1.152.2%measured
20GPT-6.2 Astra52.1%estimated ± 4.3 pp, medium confidence
21Qwen4 Max51.7%measured
22Gemini 3.8 Flash51.4%estimated ± 4.9 pp, high confidence
23Kimi K3.1 Pro51.1%measured
24Gemini 3.7 Flash51.1%estimated ± 4.9 pp, high confidence
25Gemini 3 Pro51.1%estimated ± 4.9 pp, high confidence
26Muse Wonder50.7%estimated ± 4.9 pp, high confidence
27GPT-6.2 Sol50.3%estimated ± 4.3 pp, medium confidence
28GPT-6 Astra50.3%estimated ± 4.3 pp, medium confidence
29Sakana Fugu-Ultra50.0%measured
30Claude Opus 4.849.8%measured
31Ling 3.0 Flash Sante49.8%estimated ± 4.9 pp, medium confidence
32Hy449.2%estimated ± 6.0 pp, medium confidence
33Pareto 26.949.0%measured
34Grok 4.648.6%estimated ± 4.9 pp, high confidence
35Gemini 3.6 Flash48.2%estimated ± 4.9 pp, high confidence
36Grok 4.547.8%estimated ± 4.9 pp, high confidence
37GPT-5.6 Sol47.4%estimated ± 4.3 pp, high confidence
38Claude Opus 4.6 (Adaptive)47.4%estimated ± 4.9 pp, high confidence
39Sakana Fugu47.2%measured
40DeepSeek V4 Pro47.0%estimated ± 2.7 pp, medium confidence
41Claude Opus 4.746.9%measured
42ERNIE 5.5 Flash Preview46.4%measured
43Hy4 flash46.1%estimated ± 6.0 pp, medium confidence
44Claude Haiku 5.545.9%measured
45Gemini 3.5 Flash45.8%estimated ± 2.7 pp, medium confidence
46Sage-A1.545.5%estimated ± 7.7 pp, medium confidence
47Gemini 3.1 Pro45.4%measured
48Gemini 3 Flash45.3%estimated ± 4.9 pp, high confidence
49Muse Ember44.7%estimated ± 4.9 pp, high confidence
50Ornith-1.5-397B44.6%measured
51Grok 4.744.3%estimated ± 4.9 pp, high confidence
52Inkling 2 Small44.1%measured
53dots3-note Preview44.1%estimated ± 6.0 pp, medium confidence
54Muse Spark 1.243.9%estimated ± 4.9 pp, high confidence
55Qwen3.8 Max43.6%measured
56Kimi K343.5%measured
57Hy4 preview43.4%measured
58MiniMax M3.1 Flash43.4%measured
59GPT-5.6 Terra43.4%estimated ± 4.3 pp, high confidence
60Claude Sonnet 543.2%measured
61GPT-5.5 Pro43.1%measured
62Qwen3.8 Max Preview42.8%estimated ± 4.9 pp, high confidence
63Muse Spark42.8%measured
64GPT-5.4 Pro42.7%measured
65GPT-5.3 Codex42.3%estimated ± 5.4 pp, medium confidence
66DeepSeek V4.1 Flash 101742.3%estimated ± 2.7 pp, high confidence
67Pareto 26.10 Preview42.0%estimated ± 4.3 pp, high confidence
68GPT-5.6 Luna41.8%estimated ± 4.3 pp, high confidence
69GPT-5.541.4%measured
70Qwen3.6 Plus41.1%estimated ± 4.9 pp, high confidence
71GLM-5.240.5%measured
72Agents-A140.1%estimated ± 6.0 pp, medium confidence
73Claude Opus 4.640.0%measured
74GPT-5.439.8%measured
75Claude Sonnet 4.5 Thinking39.7%estimated ± 4.9 pp, high confidence
76Step 5 Preview39.4%estimated ± 2.7 pp, high confidence
77Claude Opus 4.5 Thinking39.2%estimated ± 4.9 pp, high confidence
78Claude Sonnet 4.639.2%estimated ± 4.9 pp, high confidence
79Hy3 Preview39.2%estimated ± 2.7 pp, medium confidence
80MiMo-V2-Pro38.8%estimated ± 4.9 pp, high confidence
81GPT-5.2-Codex38.7%estimated ± 4.9 pp, high confidence
82Qwen 3.6 Max (preview)38.6%estimated ± 4.9 pp, high confidence
83Qwen3.7 Max38.6%estimated ± 2.7 pp, high confidence
84DeepSeek V4 Flash 073138.5%estimated ± 2.7 pp, high confidence
85GPT-6 Luna38.2%estimated ± 4.9 pp, high confidence
86Qwen3.8-Omni-Flash38.2%estimated ± 4.3 pp, high confidence
87MiMo-V2.6-Flash37.3%estimated ± 4.9 pp, high confidence
88GLM-5-Turbo37.2%estimated ± 4.9 pp, high confidence
89GLM-5.337.0%estimated ± 4.9 pp, high confidence
90Mistral Large 436.8%estimated ± 4.9 pp, high confidence
91Particle-136.4%estimated ± 4.3 pp, high confidence
92Qwen3.7 Plus36.1%estimated ± 4.3 pp, high confidence
93Kimi K2.636.0%estimated ± 2.7 pp, high confidence
94Beam 1.0 501B-A23B36.0%measured
95Beam 1.0 501B-A23B FP836.0%measured
96Qwen3.8-Flash-Next35.9%measured
97Muse Spark Open35.8%estimated ± 4.9 pp, high confidence
98GPT-535.7%estimated ± 4.9 pp, high confidence
99GPT-5.135.3%estimated ± 4.9 pp, high confidence
100MiMo-V2-Omni35.0%estimated ± 4.9 pp, high confidence
101Interfaze Beta35.0%estimated ± 4.3 pp, high confidence
102GLM-534.9%estimated ± 2.7 pp, medium confidence
103GPT-5.1-Codex34.8%estimated ± 4.9 pp, high confidence
104GPT-5.1-Codex-Max34.8%estimated ± 4.9 pp, high confidence
105Beam 1.0 501B-A23B NVFP434.8%measured
106GLM-5V-Turbo34.7%estimated ± 4.9 pp, high confidence
107Kimi K2.7 Code34.5%estimated ± 5.4 pp, medium confidence
108Gemini 3.1 Flash-Lite34.4%estimated ± 4.9 pp, high confidence
109GPT-5.234.4%estimated ± 4.9 pp, high confidence
110GLM-5.134.3%estimated ± 2.7 pp, medium confidence
111Qwen3.5-27B34.2%estimated ± 4.9 pp, high confidence
112GLM-5.3-Flash34.1%estimated ± 4.9 pp, high confidence
113MiMo-V2.5-Pro34.0%measured
114Command A+33.9%estimated ± 4.9 pp, high confidence
115o3-pro33.4%estimated ± 4.9 pp, high confidence
116Kimi K2.533.3%estimated ± 2.7 pp, high confidence
117Mercury 333.0%estimated ± 4.3 pp, high confidence
118Qwen3.6-27B33.0%estimated ± 4.9 pp, high confidence
119Hy333.0%estimated ± 5.4 pp, medium confidence
120Gemini 3.5 Flash-Lite32.9%estimated ± 4.9 pp, high confidence
121Grok 4.332.9%estimated ± 4.9 pp, high confidence
122MiniMax M2.732.5%estimated ± 2.7 pp, high confidence
123Claude 4 Opus Thinking32.4%estimated ± 4.9 pp, high confidence
124Solar Pro 432.4%estimated ± 4.3 pp, high confidence
125Claude Opus 4.532.3%estimated ± 4.9 pp, high confidence
126o332.3%estimated ± 4.9 pp, high confidence
127Nemotron 3.5 Super VL 120B A12B31.8%measured
128Grok 4.2031.6%measured
129Inkling-Small31.6%measured
130Step 3.7 Flash31.4%estimated ± 4.9 pp, high confidence
131Nemotron 3.5 Super VL 120B A12B NVFP431.4%measured
132Qwen3.5-35B-A3B31.3%estimated ± 4.9 pp, high confidence
133Grok 431.3%estimated ± 4.9 pp, high confidence
134Qwen3.8-27B30.8%measured
135Claude 4.1 Opus30.7%estimated ± 4.9 pp, high confidence
136Inkling 2 Mini30.7%measured
137OrcaCyber-Nano-1.030.3%estimated ± 4.3 pp, high confidence
138Inkling30.0%measured
139Ling 3.0 Flash29.9%estimated ± 2.7 pp, medium confidence
140Claude 4 Opus29.1%estimated ± 4.9 pp, high confidence
141Claude 4 Sonnet29.1%estimated ± 4.9 pp, high confidence
142Qwen3.5 397B29.1%estimated ± 5.3 pp, medium confidence
143Qwen3 Max28.9%estimated ± 4.9 pp, high confidence
144Solar Open 228.8%measured
145Nemotron 3.5 Bolt 74B-A9B BF1628.7%measured
146Gemini 2.5 Pro28.7%estimated ± 4.9 pp, high confidence
147MiMo-V2-Flash28.7%estimated ± 4.9 pp, high confidence
148DeepSeek V3.228.7%estimated ± 4.9 pp, high confidence
149Grok 4.1 Fast (Reasoning)28.5%estimated ± 4.9 pp, high confidence
150MiniMax M328.5%estimated ± 4.9 pp, high confidence
151Qwen3.5 Flash28.3%estimated ± 4.9 pp, high confidence
152GPT-5.4 mini28.2%measured
153K-Exaone27.4%estimated ± 4.9 pp, high confidence
154o127.3%estimated ± 4.9 pp, high confidence
155Grok Code Fast 127.1%estimated ± 4.9 pp, high confidence
156Ling 2.6 Flash27.1%estimated ± 4.9 pp, high confidence
157DeepSeek V3.126.8%estimated ± 4.9 pp, high confidence
158Apodex 1.126.8%estimated ± 4.9 pp, high confidence
159Apodex 1.1 Mini26.8%estimated ± 4.9 pp, high confidence
160Nemotron 3 Ultra26.7%measured
161MiMo-V2.526.5%estimated ± 4.9 pp, high confidence
162A.X K226.5%estimated ± 2.7 pp, medium confidence
163DeepSeek-R126.4%estimated ± 4.9 pp, high confidence
164Claude 3.7 Sonnet Thinking26.3%estimated ± 4.9 pp, high confidence
165GLM-4.726.3%estimated ± 4.9 pp, high confidence
166Kimi K226.0%estimated ± 4.9 pp, high confidence
167GLM-4.625.8%estimated ± 4.9 pp, high confidence
168o1-pro25.8%estimated ± 4.9 pp, high confidence
169Ornith-1.5-35B-A3B25.6%measured
170Nemotron 3.5 Bolt 74B-A9B25.3%measured
171GLM-4.525.2%estimated ± 4.9 pp, high confidence
172Claude 3.7 Sonnet25.0%estimated ± 4.9 pp, high confidence
173o4-mini25.0%estimated ± 4.9 pp, high confidence
174o1-preview25.0%estimated ± 4.9 pp, high confidence
175GPT-4.124.9%estimated ± 4.9 pp, high confidence
176Grok 4.1 Fast24.9%estimated ± 4.9 pp, high confidence
177Mistral Large 324.7%estimated ± 4.9 pp, high confidence
178GLM-4.5-Air24.7%estimated ± 4.9 pp, high confidence
179Grok 4 Fast (Reasoning)24.7%estimated ± 4.9 pp, high confidence
180Claude Haiku 4.524.6%estimated ± 4.9 pp, high confidence
181o3-mini24.6%estimated ± 4.9 pp, high confidence
182Claude 3.5 Sonnet24.5%estimated ± 4.9 pp, high confidence
183Gemini 2.0 Flash24.4%estimated ± 4.9 pp, high confidence
184GPT-4.1 mini24.4%estimated ± 4.9 pp, high confidence
185Gemini 1.5 Pro24.4%estimated ± 4.9 pp, medium confidence
186Mistral Medium 3.5 128B24.4%estimated ± 4.9 pp, medium confidence
187GPT-4o24.3%estimated ± 4.9 pp, medium confidence
188Laguna XS.224.3%estimated ± 4.9 pp, medium confidence
189Laguna M.124.3%estimated ± 4.9 pp, medium confidence
190Gemini 1.5 Flash24.3%estimated ± 4.9 pp, medium confidence
191Claude 3.5 Haiku24.3%estimated ± 4.9 pp, medium confidence
192GPT-4.1 nano24.3%estimated ± 4.9 pp, medium confidence
193GPT-4o mini24.3%estimated ± 4.9 pp, medium confidence
194GPT-5.4 nano24.3%measured
195Qwen3.5-122B-A10B23.8%estimated ± 5.3 pp, medium confidence
196Gemini 2.5 Flash23.7%estimated ± 4.9 pp, high confidence
197Ternary Bonsai 2 27B23.7%estimated ± 4.3 pp, high confidence
198o1-mini23.7%estimated ± 4.9 pp, high confidence
199DeepSeek V3 032423.6%estimated ± 4.9 pp, high confidence
200GPT-4.523.5%estimated ± 4.9 pp, high confidence
201QwQ-32B23.4%estimated ± 4.9 pp, high confidence
202Magistral Medium23.1%estimated ± 4.9 pp, high confidence
203Mistral Medium 323.1%estimated ± 4.9 pp, high confidence
204K-EXAONE 2.023.1%estimated ± 2.7 pp, medium confidence
205Claude 3 Opus22.8%estimated ± 4.9 pp, high confidence
206DeepSeek V322.6%estimated ± 4.9 pp, high confidence
207DeepSeek R1 Distill Qwen 32B22.6%estimated ± 4.9 pp, high confidence
208Magistral Small22.4%estimated ± 4.9 pp, high confidence
209Ripple-65B Preview22.4%estimated ± 4.3 pp, high confidence
210Qwen3.6-35B-A3B22.3%estimated ± 5.3 pp, medium confidence
211Solar Pro 322.1%estimated ± 4.9 pp, high confidence
212Devstral Small21.9%estimated ± 4.9 pp, high confidence
213Mistral Large 221.9%estimated ± 4.9 pp, high confidence
214Mistral Small 3.121.5%estimated ± 4.9 pp, high confidence
215GPT-4 Turbo21.5%estimated ± 4.9 pp, high confidence
216Solar Pro 221.4%estimated ± 4.9 pp, high confidence
217Nova Pro21.4%estimated ± 4.9 pp, high confidence
218Qwen2.5 Coder 32B Instruct21.2%estimated ± 4.9 pp, high confidence
219Gemini 2.5 Flash-Lite21.2%estimated ± 4.9 pp, high confidence
220DeepSeek-V2.521.1%estimated ± 4.9 pp, high confidence
221Ling 3.0 Flash VL21.0%estimated ± 5.4 pp, medium confidence
222Muse Glimmer 30B20.9%estimated ± 4.9 pp, high confidence
223Claude 3 Haiku20.3%estimated ± 4.9 pp, high confidence
224Granite 4.2 30B20.3%estimated ± 4.9 pp, high confidence
225MAI-Thinking-120.3%estimated ± 4.3 pp, high confidence
226Ornith-1.5-9B20.2%measured
227Gemini 1.0 Pro20.1%estimated ± 4.9 pp, high confidence
228Granite 4.2 8B19.9%estimated ± 4.9 pp, high confidence
229Ling 3.0 Flash FP819.9%estimated ± 4.3 pp, high confidence
230Gemma 3 27B19.7%estimated ± 4.9 pp, high confidence
231Gemma 4 E4B19.7%estimated ± 4.9 pp, high confidence
232Granite 4.2 3B19.6%estimated ± 4.9 pp, high confidence
233Gemma 4 31B19.5%measured
234Mistral Small 419.3%estimated ± 4.9 pp, medium confidence
235GPT-OSS 120B18.9%estimated ± 4.9 pp, medium confidence
236North Mini Code18.6%estimated ± 4.9 pp, medium confidence
237Spark-X2.6-4B17.9%measured
238Quasar 438B17.6%estimated ± 5.4 pp, medium confidence
239Claude Sonnet 4.517.3%estimated ± 5.3 pp, medium confidence
240DeepSeek-R1-052814.6%estimated ± 5.3 pp, medium confidence
241Grok 3 Mini14.0%estimated ± 5.3 pp, medium confidence
242Nemotron 3 Super 100B13.4%estimated ± 5.3 pp, medium confidence
243Mercury 2.512.9%estimated ± 4.3 pp, high confidence
244Sarvam 105B12.7%estimated ± 4.3 pp, high confidence
245Qwen3 235B 250712.6%estimated ± 5.3 pp, medium confidence
246Nemotron Ultra 253B12.2%estimated ± 5.3 pp, medium confidence
247Grok 3 [Beta]12.0%estimated ± 5.3 pp, medium confidence
248GLM-4.7-Flash12.0%estimated ± 5.3 pp, medium confidence
249Nemotron 3 Nano 30B11.7%estimated ± 5.3 pp, medium confidence
250GPT-OSS 20B11.5%estimated ± 5.3 pp, medium confidence
251Qwen3 235B-A22B (Reasoning)11.5%estimated ± 5.3 pp, medium confidence
252Phi-4-reasoning-plus11.4%estimated ± 5.3 pp, low confidence
253Qwen3 32B (Reasoning)11.4%estimated ± 5.3 pp, low confidence
254Phi-4-reasoning11.3%estimated ± 5.3 pp, low confidence
255DeepSeek R1 Distill Llama 70B11.3%estimated ± 5.3 pp, low confidence
256Qwen3 235B-A22B11.3%estimated ± 5.3 pp, low confidence
257DeepSeek-R1-0528-Qwen3-8B11.3%estimated ± 5.3 pp, low confidence
258Qwen2.5-Max11.3%estimated ± 5.3 pp, low confidence
259DeepSeek R1 Distill Qwen 14B11.3%estimated ± 5.3 pp, low confidence
260Phi-411.3%estimated ± 5.3 pp, low confidence
261Grok 211.3%estimated ± 5.3 pp, low confidence
262Qwen3 32B11.3%estimated ± 5.3 pp, low confidence
263MiniMax-Text-0111.3%estimated ± 5.3 pp, low confidence
264Grok 2 mini11.3%estimated ± 5.3 pp, low confidence
265Gemini 2.0 Flash-Lite11.3%estimated ± 5.3 pp, low confidence
266Llama 3.3 70B11.3%estimated ± 5.3 pp, low confidence
267Llama 3.1 70B11.3%estimated ± 5.3 pp, low confidence
268Qwen2.5-VL-32B11.3%estimated ± 5.3 pp, low confidence
269Trinity-Large-Thinking11.1%estimated ± 4.3 pp, high confidence
270Nemotron 3.5 Lightning 30B A3B NVFP410.5%measured
271LFM2.5-24B-A2B10.3%estimated ± 4.3 pp, medium confidence
272Nemotron 3 Nano Omni 30B A3B9.6%estimated ± 4.3 pp, medium confidence
273ZAYA1-8B9.4%estimated ± 4.3 pp, medium confidence
274MiniCPM5-2B9.3%estimated ± 4.3 pp, medium confidence
275LongCat-Flash-Lite-Sparse9.2%estimated ± 4.3 pp, medium confidence
276Sarvam 30B8.9%estimated ± 4.3 pp, medium confidence
277Mellum2.1-12B-A2.5B-Thinking8.8%estimated ± 4.3 pp, medium confidence
278LFM3-1.7B8.8%estimated ± 4.3 pp, medium confidence
279Trinity-Large-Preview8.8%estimated ± 4.3 pp, medium confidence
280Gemma 4 26B A4B8.7%measured
281Mellum2-12B-A2.5B-Thinking8.7%estimated ± 4.3 pp, medium confidence
282ZAYA1-74B-Preview8.7%estimated ± 4.3 pp, medium confidence
283Qwen3.5-4B8.5%estimated ± 5.4 pp, low confidence
284Ling 3.0 Tiny7.9%estimated ± 5.4 pp, low confidence
285Qwen3-Omni-30B-A3B-Thinking6.0%estimated ± 5.4 pp, low confidence
286LFM2.5-8B-A1B5.4%estimated ± 5.4 pp, low confidence
287Celeris-15.3%estimated ± 5.4 pp, low confidence
288Gemma 4 12B5.2%measured
289Granite-4.0-H-350M4.9%estimated ± 5.4 pp, low confidence
290LFM2.5-2.6B4.7%estimated ± 5.4 pp, low confidence
291Exaone 4.0 1.2B4.2%estimated ± 5.4 pp, low confidence
292Granite-4.0-350M4.0%estimated ± 5.4 pp, low confidence
293LFM2.5-VL-1.6B-Extract3.5%estimated ± 5.4 pp, low confidence
294Exaone 4.0 32B3.4%estimated ± 5.4 pp, low confidence
295Granite-4.0-H-1B3.4%estimated ± 5.4 pp, low confidence
296Phi-4 Multimodal Instruct3.4%estimated ± 5.4 pp, low confidence
297Llama 4 Maverick3.3%estimated ± 5.4 pp, low confidence
298Gemma 4 E2B3.2%estimated ± 5.4 pp, low confidence
299Qwen3-Omni-30B-A3B-Instruct3.0%estimated ± 5.4 pp, low confidence
300Llama 3.1 8B2.7%estimated ± 5.4 pp, low confidence
301Llama 3.1 405B2.4%estimated ± 5.4 pp, low confidence
302Llama 4 Scout2.2%estimated ± 5.4 pp, low confidence
303Ultravox v0.6 Llama 3.3 70B2.0%estimated ± 5.4 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General