Knowledge & reasoning

HLE leaderboard

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

As of 2026-10-07, the highest measured score on HLE is 65.0% by Claude Fable 5.1. 185 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 5.5	67.6%	estimated ± 1.9 pp, high confidence
2	Claude Fable 5.1	65.0%	measured
3	Claude Opus 5	64.7%	measured
4	Claude Mythos 5	64.5%	measured
5	Claude Sonnet 5.5	63.7%	estimated ± 5.7 pp, medium confidence
6	Muse Spark 1.1	62.1%	measured
7	Claude Fable 5	61.7%	estimated ± 6.3 pp, medium confidence
8	GPT-6 Astra	60.9%	estimated ± 5.5 pp, low confidence
9	GPT-6.1 Sol	60.9%	estimated ± 5.5 pp, low confidence
10	Gemini 3.8 Flash	60.7%	estimated ± 5.5 pp, low confidence
11	GPT-5.6 Sol	60.7%	estimated ± 5.5 pp, low confidence
12	GPT-6 Sol	60.5%	estimated ± 5.5 pp, low confidence
13	Grok 4.6	60.0%	estimated ± 5.5 pp, low confidence
14	GPT-5.4 Pro	58.7%	measured
15	GPT-5.6 Terra	58.6%	estimated ± 5.5 pp, low confidence
16	Sakana Fugu-Ultra	58.3%	estimated ± 5.7 pp, medium confidence
17	Claude Opus 4.8	57.9%	measured
18	Gemini 3.7 Flash	57.8%	estimated ± 6.3 pp, medium confidence
19	Pareto 26.9	57.4%	estimated ± 5.7 pp, medium confidence
20	Claude Sonnet 5	57.4%	measured
21	GPT-5.5 Pro	57.2%	measured
22	Gemini 3.1 Pro	56.6%	estimated ± 5.5 pp, low confidence
23	Apodex 1.1	56.1%	measured
24	Kimi K3	56.0%	measured
25	Sakana Fugu	55.9%	estimated ± 5.7 pp, medium confidence
26	Gemini 4 Argon	55.5%	estimated ± 9.5 pp, medium confidence
27	Hy4 preview	55.4%	measured
28	MiMo-V2.6-Pro	55.2%	estimated ± 9.5 pp, medium confidence
29	Grok 4.5	55.0%	estimated ± 5.5 pp, low confidence
30	Claude Haiku 5.5	54.8%	estimated ± 5.7 pp, medium confidence
31	Claude Opus 4.7 (Adaptive)	54.7%	measured
32	GLM-5.2	54.7%	measured
33	Muse Spark 1.3	54.4%	estimated ± 9.5 pp, medium confidence
34	Claude Opus 4.6	53.0%	measured
35	dots3-note Preview	52.6%	measured
36	GLM-5.1	52.3%	measured
37	GPT-5.5	52.2%	measured
38	GPT-5.4	52.1%	measured
39	GPT-5.6 Luna	51.6%	estimated ± 5.5 pp, low confidence
40	GLM-5.3	51.1%	estimated ± 9.5 pp, medium confidence
41	GLM-5	50.4%	measured
42	Muse Spark	50.4%	measured
43	Ling 3.1 Flash	50.4%	estimated ± 9.5 pp, medium confidence
44	Grok 4.7	50.2%	estimated ± 9.5 pp, medium confidence
45	Muse Spark 1.2	50.2%	estimated ± 9.5 pp, medium confidence
46	Qwen3.8 Max Preview	50.2%	estimated ± 9.5 pp, medium confidence
47	GPT-5.3 Codex	49.6%	estimated ± 9.5 pp, medium confidence
48	Claude Sonnet 4.6	49.0%	measured
49	GLM-5.3-Flash	48.6%	estimated ± 9.5 pp, medium confidence
50	Claude Opus 4.6 (Adaptive)	48.5%	estimated ± 6.3 pp, medium confidence
51	MiMo-V2.5-Pro	48.0%	measured
52	Inkling-Small	47.8%	measured
53	Agents-A1	47.6%	measured
54	Pareto 26.10 Preview	47.0%	estimated ± 9.2 pp, medium confidence
55	Step 5 Preview	46.5%	measured
56	Gemini 3.6 Flash	46.0%	estimated ± 6.3 pp, medium confidence
57	Inkling	46.0%	measured
58	GPT-6 Luna	46.0%	estimated ± 5.5 pp, low confidence
59	Beam	45.9%	estimated ± 5.7 pp, medium confidence
60	MiMo-V2.6-Flash	45.8%	estimated ± 9.5 pp, medium confidence
61	GPT-5.2	44.8%	estimated ± 6.3 pp, medium confidence
62	Grok 4.20	44.8%	estimated ± 5.5 pp, low confidence
63	Ornith-1.5-397B	44.6%	measured
64	Mistral Large 4	44.4%	estimated ± 9.5 pp, medium confidence
65	Gemini 3 Pro Deep Think	44.2%	estimated ± 6.3 pp, medium confidence
66	Claude Opus 4.5 Thinking	44.0%	estimated ± 6.3 pp, medium confidence
67	Gemini 3 Pro	43.9%	estimated ± 6.3 pp, medium confidence
68	Claude Sonnet 4.5	43.8%	estimated ± 6.3 pp, medium confidence
69	Claude Sonnet 4.5 Thinking	43.8%	estimated ± 6.3 pp, medium confidence
70	Kimi K2.7 Code	43.8%	estimated ± 9.5 pp, medium confidence
71	Qwen3.8 Max	43.6%	measured
72	Gemini 3.1 Flash-Lite	43.5%	estimated ± 9.8 pp, low confidence
73	Quasar 438B	43.2%	estimated ± 9.5 pp, medium confidence
74	DeepSeek V4 Pro 0813	42.7%	measured
75	GPT-5.2-Codex	42.4%	estimated ± 9.5 pp, medium confidence
76	GPT-5.1	42.2%	estimated ± 6.4 pp, medium confidence
77	GPT-5.4 mini	41.5%	measured
78	Qwen3.7 Max	41.4%	measured
79	Gemini 3.5 Flash	40.2%	measured
80	Qwen3.5 Flash	39.5%	estimated ± 9.8 pp, low confidence
81	GPT-5.1-Codex	38.5%	estimated ± 9.5 pp, medium confidence
82	GPT-5.1-Codex-Max	38.5%	estimated ± 9.5 pp, medium confidence
83	GPT-5 (high)	38.5%	estimated ± 9.5 pp, medium confidence
84	Solar Open 2	38.3%	estimated ± 5.7 pp, medium confidence
85	Interfaze Beta	37.9%	estimated ± 8.2 pp, medium confidence
86	Solar Pro 4	37.8%	estimated ± 9.2 pp, medium confidence
87	GPT-5.4 nano	37.7%	measured
88	Claude Opus 4.7	37.5%	estimated ± 9.5 pp, medium confidence
89	MiMo-V2.5	37.3%	estimated ± 9.8 pp, low confidence
90	Hy3	37.2%	estimated ± 9.5 pp, medium confidence
91	DeepSeek V4.1 Flash	36.8%	measured
92	Apodex 1.1 Mini	36.7%	estimated ± 9.5 pp, medium confidence
93	Qwen3.8-Omni-Flash	36.5%	measured
94	Qwen3.8-Flash-Next	35.9%	measured
95	Qwen 3.6 Max (preview)	35.5%	estimated ± 7.8 pp, medium confidence
96	Grok 4.3	35.0%	measured
97	MiniMax M3	35.0%	estimated ± 9.5 pp, medium confidence
98	DeepSeek V4 Flash 0731	34.8%	measured
99	Kimi K2.6	34.7%	measured
100	Qwen3.7 Plus	34.7%	measured
101	GLM-4.5	34.1%	estimated ± 9.8 pp, low confidence
102	Nemotron 3 Super 100B	33.7%	estimated ± 9.5 pp, medium confidence
103	MiniMax M2.7	33.3%	estimated ± 9.2 pp, medium confidence
104	Grok 4.1 Fast (Reasoning)	33.2%	estimated ± 9.5 pp, medium confidence
105	Grok 4 Fast (Reasoning)	33.2%	estimated ± 9.5 pp, medium confidence
106	Seed 2.1 Pro	33.1%	estimated ± 7.8 pp, medium confidence
107	Muse Glimmer 30B	32.5%	estimated ± 9.5 pp, medium confidence
108	Step 3.7 Flash	31.7%	estimated ± 9.5 pp, medium confidence
109	Kimi K2.5 (Reasoning)	31.7%	estimated ± 8.2 pp, medium confidence
110	DeepSeek V3.1 (Reasoning)	30.9%	estimated ± 9.5 pp, medium confidence
111	Grok 4	30.9%	estimated ± 9.5 pp, medium confidence
112	Ling 3.0 Flash VL	30.9%	estimated ± 9.5 pp, medium confidence
113	Claude Opus 4.5	30.8%	measured
114	Qwen3.8-27B	30.8%	measured
115	Seed 2.1 Turbo	30.5%	estimated ± 7.8 pp, medium confidence
116	Qwen3.5-122B-A10B	30.2%	estimated ± 7.8 pp, medium confidence
117	Kimi K2.5	30.1%	measured
118	Claude Haiku 4.5	29.8%	estimated ± 9.8 pp, low confidence
119	Qwen3.5-27B	29.1%	estimated ± 7.8 pp, medium confidence
120	DeepSeek-R1	29.0%	estimated ± 9.5 pp, medium confidence
121	Gemini 2.5 Flash	29.0%	estimated ± 9.5 pp, medium confidence
122	Gemini 3 Flash	29.0%	estimated ± 9.5 pp, medium confidence
123	GPT-OSS 20B	29.0%	estimated ± 9.5 pp, medium confidence
124	Qwen3.6 Plus	28.8%	measured
125	Qwen3.5 397B	28.7%	measured
126	Claude 4 Sonnet	27.8%	estimated ± 9.5 pp, medium confidence
127	GPT-OSS 120B	27.8%	estimated ± 9.5 pp, medium confidence
128	K-Exaone	27.8%	estimated ± 9.5 pp, medium confidence
129	MiMo-V2-Omni	27.8%	estimated ± 9.5 pp, medium confidence
130	o3	27.8%	estimated ± 9.5 pp, medium confidence
131	A.X K2	27.8%	measured
132	o3-pro	27.5%	estimated ± 9.9 pp, medium confidence
133	Qwen3.5-35B-A3B	27.4%	estimated ± 7.8 pp, medium confidence
134	DeepSeek V3.2	27.0%	estimated ± 9.5 pp, medium confidence
135	Nemotron 3 Nano 30B	27.0%	estimated ± 9.5 pp, medium confidence
136	Qwen3.5 397B (Reasoning)	27.0%	estimated ± 9.5 pp, medium confidence
137	Qwen3 235B 2507	26.8%	estimated ± 7.8 pp, medium confidence
138	Nemotron 3 Ultra	26.7%	measured
139	Gemma 4 31B	26.5%	measured
140	Ornith-1.5-35B-A3B	25.6%	measured
141	Hy3 Preview	25.5%	measured
142	GLM-5V-Turbo	25.5%	estimated ± 9.5 pp, medium confidence
143	MAI-Thinking-1	25.1%	estimated ± 8.2 pp, medium confidence
144	Ling 3.0 Flash FP8	24.8%	estimated ± 8.2 pp, medium confidence
145	GLM-4.7	24.8%	measured
146	MiMo-V2-Flash	24.4%	estimated ± 8.2 pp, medium confidence
147	Qwen3.6-27B	24.0%	measured
148	Command A+	23.5%	estimated ± 9.5 pp, medium confidence
149	GLM-5-Turbo	23.5%	estimated ± 9.5 pp, medium confidence
150	MiMo-V2-Pro	23.5%	estimated ± 9.5 pp, medium confidence
151	Mistral Small 4	23.5%	estimated ± 9.5 pp, medium confidence
152	Mistral Small 4 (Reasoning)	23.5%	estimated ± 9.5 pp, medium confidence
153	North Mini Code	23.5%	estimated ± 9.5 pp, medium confidence
154	Sarvam 30B	23.5%	estimated ± 9.5 pp, medium confidence
155	Ling 3.0 Flash	22.7%	measured
156	Qwen3.6-35B-A3B	21.4%	measured
157	Claude 4.1 Opus	20.3%	estimated ± 9.9 pp, medium confidence
158	Ornith-1.5-9B	20.2%	measured
159	o1-preview	19.4%	estimated ± 9.9 pp, low confidence
160	Claude 3 Opus	19.4%	estimated ± 9.9 pp, low confidence
161	DeepSeek R1 Distill Qwen 32B	19.4%	estimated ± 9.9 pp, low confidence
162	Gemini 1.5 Pro	19.4%	estimated ± 9.9 pp, low confidence
163	GPT-4 Turbo	19.4%	estimated ± 9.9 pp, low confidence
164	GPT-4o mini	19.4%	estimated ± 9.9 pp, low confidence
165	Phi-4 Multimodal Instruct	19.4%	estimated ± 9.9 pp, low confidence
166	Qwen2.5 Coder 32B Instruct	19.4%	estimated ± 9.9 pp, low confidence
167	Gemini 1.0 Pro	19.4%	estimated ± 9.9 pp, low confidence
168	Mercury 2.5	19.2%	estimated ± 9.2 pp, medium confidence
169	Celeris-1	19.0%	estimated ± 9.5 pp, medium confidence
170	Claude 3 Haiku	19.0%	estimated ± 9.5 pp, medium confidence
171	Claude 4.1 Opus Thinking	19.0%	estimated ± 9.5 pp, medium confidence
172	DeepSeek V3 0324	19.0%	estimated ± 9.5 pp, medium confidence
173	DeepSeek V3.1	19.0%	estimated ± 9.5 pp, medium confidence
174	Exaone 4.0 1.2B	19.0%	estimated ± 9.5 pp, medium confidence
175	Exaone 4.0 32B	19.0%	estimated ± 9.5 pp, medium confidence
176	Gemini 3.5 Flash-Lite	19.0%	estimated ± 9.5 pp, medium confidence
177	Gemma 3 27B	19.0%	estimated ± 9.5 pp, medium confidence
178	GLM-4.5-Air	19.0%	estimated ± 9.5 pp, medium confidence
179	GLM-4.6	19.0%	estimated ± 9.5 pp, medium confidence
180	GPT-4o	19.0%	estimated ± 9.5 pp, medium confidence
181	GPT-5 (medium)	19.0%	estimated ± 9.5 pp, medium confidence
182	Granite-4.0-350M	19.0%	estimated ± 9.5 pp, medium confidence
183	Granite-4.0-H-1B	19.0%	estimated ± 9.5 pp, medium confidence
184	Granite-4.0-H-350M	19.0%	estimated ± 9.5 pp, medium confidence
185	Grok 4.1 Fast	19.0%	estimated ± 9.5 pp, medium confidence
186	Grok Code Fast 1	19.0%	estimated ± 9.5 pp, medium confidence
187	Kimi K2	19.0%	estimated ± 9.5 pp, medium confidence
188	LFM2.5-2.6B	19.0%	estimated ± 9.5 pp, medium confidence
189	LFM2.5-8B-A1B	19.0%	estimated ± 9.5 pp, medium confidence
190	LFM2.5-VL-1.6B-Extract	19.0%	estimated ± 9.5 pp, medium confidence
191	Ling 3.0 Tiny	19.0%	estimated ± 9.5 pp, medium confidence
192	Llama 3.1 405B	19.0%	estimated ± 9.5 pp, medium confidence
193	Llama 4 Maverick	19.0%	estimated ± 9.5 pp, medium confidence
194	Llama 4 Scout	19.0%	estimated ± 9.5 pp, medium confidence
195	Mistral Large 2	19.0%	estimated ± 9.5 pp, medium confidence
196	Mistral Large 3	19.0%	estimated ± 9.5 pp, medium confidence
197	Mistral Medium 3	19.0%	estimated ± 9.5 pp, medium confidence
198	Mistral Medium 3.5 128B	19.0%	estimated ± 9.5 pp, medium confidence
199	Nemotron Ultra 253B	19.0%	estimated ± 9.5 pp, medium confidence
200	Nova Pro	19.0%	estimated ± 9.5 pp, medium confidence
201	Phi-4	19.0%	estimated ± 9.5 pp, medium confidence
202	Qwen3 Max	19.0%	estimated ± 9.5 pp, medium confidence
203	Qwen3-Omni-30B-A3B-Instruct	19.0%	estimated ± 9.5 pp, medium confidence
204	Qwen3-Omni-30B-A3B-Thinking	19.0%	estimated ± 9.5 pp, medium confidence
205	Sarvam 105B	19.0%	estimated ± 9.5 pp, medium confidence
206	Solar Pro 2	19.0%	estimated ± 9.5 pp, medium confidence
207	Solar Pro 3	19.0%	estimated ± 9.5 pp, medium confidence
208	Ultravox v0.6 Llama 3.3 70B	19.0%	estimated ± 9.5 pp, medium confidence
209	Gemini 2.5 Pro	18.8%	measured
210	Ternary Bonsai 2 27B	18.7%	estimated ± 5.9 pp, low confidence
211	o1-pro	18.6%	estimated ± 8.2 pp, low confidence
212	K-EXAONE 2.0	18.3%	measured
213	Gemma 4 26B A4B	17.2%	measured
214	o3-mini	17.0%	estimated ± 8.2 pp, low confidence
215	Nemotron 3.5 Lightning 30B A3B NVFP4	15.8%	estimated ± 5.7 pp, medium confidence
216	o1	15.8%	estimated ± 8.2 pp, low confidence
217	Laguna XS.2	15.7%	estimated ± 9.8 pp, low confidence
218	Trinity-Large-Thinking	15.6%	estimated ± 9.2 pp, medium confidence
219	Laguna M.1	15.4%	estimated ± 9.8 pp, low confidence
220	Nemotron 3 Nano Omni 30B A3B	13.4%	estimated ± 8.2 pp, low confidence
221	ZAYA1-8B	12.7%	estimated ± 8.2 pp, low confidence
222	Mellum2-12B-A2.5B-Thinking	12.3%	estimated ± 5.9 pp, low confidence
223	LongCat-Flash-Lite-Sparse	11.1%	estimated ± 2.5 pp, low confidence
224	Granite 4.2 30B	10.4%	estimated ± 8.2 pp, low confidence
225	GPT-4.1	10.3%	estimated ± 8.2 pp, low confidence
226	GPT-4.1 mini	9.5%	estimated ± 8.2 pp, low confidence
227	Granite 4.2 8B	9.5%	estimated ± 8.2 pp, low confidence
228	MiniCPM5-2B	8.9%	measured
229	Gemma 4 12B	8.2%	estimated ± 5.7 pp, low confidence
230	Claude 3.5 Sonnet	7.8%	estimated ± 8.2 pp, low confidence
231	DeepSeek V3	7.7%	estimated ± 8.2 pp, low confidence
232	Ling 2.6 Flash	7.7%	estimated ± 8.2 pp, low confidence
233	Gemma 4 E4B	7.6%	estimated ± 8.2 pp, low confidence
234	ZAYA1-74B-Preview	7.2%	estimated ± 8.2 pp, low confidence
235	Granite 4.2 3B	6.5%	estimated ± 8.2 pp, low confidence
236	GPT-4.1 nano	5.5%	estimated ± 8.2 pp, low confidence
237	Gemma 4 E2B	4.2%	estimated ± 8.2 pp, low confidence
238	Soofi S 30B-A3B	4.2%	estimated ± 8.2 pp, low confidence
239	Trinity-Large-Preview	2.8%	estimated ± 9.2 pp, low confidence
240	LFM2.5-VL-450M	1.9%	estimated ± 8.2 pp, low confidence
241	LFM2.5-230M	1.9%	estimated ± 8.2 pp, low confidence
242	LLaDA2.2-mini	0.0%	estimated ± 9.2 pp, low confidence
243	Mellum2-12B-A2.5B-Instruct	0.0%	estimated ± 5.9 pp, low confidence
244	MiniCPM5-1B	0.0%	estimated ± 5.9 pp, low confidence

HLE leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General