Artificial Analysis Humanity's Last Exam (AA-HLE)
Data verified
An Artificial Analysis Humanity's Last Exam score.
Top models on AA-HLE — October 28, 2026
As of October 28, 2026, Claude Opus 5.5 leads the AA-HLE leaderboard with 61.4% , followed by Gemini 4 Carbon (60.4%) and Claude Fable 5.1 (59.1%).
1Closed
Claude Opus 5.5
Anthropic
61.4%
Overall 86.29Context 1M
2Closed
Gemini 4 Carbon
60.4%
Overall 88.37
3Closed
Claude Fable 5.1
Anthropic
59.1%
Overall 81.07Context 1M
241 modelsKnowledge11% of Knowledge reference weightCurrentUpdated October 28, 2026
Leaderboard (241 models)
241 of 241 rows
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 61.4% | Not reported | Closed |
| 2 | Gemini 4 CarbonGoogle | 60.4% | Not reported | Closed |
| 3 | Claude Fable 5.1Anthropic | 59.1% | Not reported | Closed |
| 4 | Gemini 4 ArgonGoogle | 57.1% | Not reported | Closed |
| 5 | Claude Fable 5Anthropic | 55.5% | Not reported | Closed |
| 6 | Claude Sonnet 5.5Anthropic | 55.0% | Not reported | Closed |
| 7 | Claude Opus 5Anthropic | 54.9% | Not reported | Closed |
| 8 | GPT-6 AstraOpenAI | 54.7% | Not reported | Closed |
| 9 | Muse Wonder FullMeta | 53.6% | Not reported | Closed |
| 10 | GPT-6.1 SolOpenAI | 52.9% | Not reported | Closed |
| 11 | Kimi K3.1 ProMoonshot AI | 52.5% | Not reported | Pending |
| 12 | Muse WonderMeta | 50.1% | Not reported | Closed |
| 13 | GPT-5.6 SolOpenAI | 49.5% | Not reported | Closed |
| 14 | MiMo-V2.6-ProXiaomi | 49.4% | Not reported | Open |
| 15 | Claude Opus 4.8Anthropic | 48.7% | Not reported | Closed |
| 16 | Muse Spark 1.3Meta | 48.7% | Not reported | Closed |
| 17 | MiniMax M3.1 FlashMiniMax | 48.0% | Not reported | Closed |
| 18 | GPT-6 SolOpenAI | 47.9% | Not reported | Closed |
| 19 | Gemini 3.7 FlashGoogle | 47.9% | Not reported | Closed |
| 20 | Gemini 3.8 FlashGoogle | 47.8% | Not reported | Closed |
| 21 | Gemini 3.1 ProGoogle | 47.0% | Not reported | Closed |
| 22 | Kimi K3Moonshot AI | 46.9% | Not reported | Pending |
| 23 | Step 5 PreviewStepFun | 46.5% | Not reported | Open |
| 24 | Muse Spark 1.1Meta | 46.2% | Not reported | Closed |
| 25 | GPT-6.1 TerraOpenAI | 46.2% | Not reported | Closed |
| 26 | GPT-5.5OpenAI | 45.8% | Not reported | Closed |
| 27 | Muse Spark 1.2Meta | 45.5% | Not reported | Closed |
| 28 | Claude Haiku 5.5Anthropic | 44.4% | Not reported | Closed |
| 29 | Muse Spark OpenMeta | 43.9% | Not reported | Open |
| 30 | GPT-5.4OpenAI | 43.7% | Not reported | Closed |
| 31 | Grok 4.7xAI | 43.1% | Not reported | Closed |
| 32 | Qwen3.8 Max PreviewAlibaba | 43.1% | Not reported | Closed |
| 33 | GPT-5.6 TerraOpenAI | 42.9% | Not reported | Closed |
| 34 | Grok 4.6xAI | 42.9% | Not reported | Closed |
| 35 | Gemini 3.5 FlashGoogle | 42.7% | Not reported | Closed |
| 36 | Grok 4.5xAI | 42.7% | Not reported | Closed |
| 37 | GPT-5.3 CodexOpenAI | 42.5% | Not reported | Closed |
| 38 | GPT-5.3-Codex-SparkOpenAI | 42.5% | Not reported | Closed |
| 39 | Claude Opus 4.7 (Adaptive)Anthropic | 42.3% | Not reported | Closed |
| 40 | GLM-5.3Z.AI | 42.3% | Not reported | Open |
| 41 | Claude Sonnet 5Anthropic | 41.3% | Not reported | Closed |
| 42 | GLM-5.2Z.AI | 41.1% | Not reported | Open |
| 43 | DeepSeek V4 Pro 0813DeepSeek | 41.0% | Not reported | Open |
| 44 | Gemini 3.6 FlashGoogle | 40.8% | Not reported | Closed |
| 45 | Muse SparkMeta | 40.7% | Not reported | Closed |
| 46 | Qwen3.7 MaxAlibaba | 40.5% | Not reported | Closed |
| 47 | Claude Opus 4.6 (Adaptive)Anthropic | 39.9% | Not reported | Closed |
| 48 | Gemini 3 ProGoogle | 39.7% | Not reported | Closed |
| 49 | GPT-5.6 LunaOpenAI | 39.5% | Not reported | Closed |
| 50 | Ling 3.1 FlashInclusionAI | 39.4% | Not reported | Open |
| 51 | DeepSeek V4.1 Flash 1017DeepSeek | 39.2% | Not reported | Open |
| 52 | MiniMax M3MiniMax | 39.0% | Not reported | Open |
| 53 | DeepSeek V4 Flash 0731DeepSeek | 38.6% | Not reported | Open |
| 54 | GPT-6 LunaOpenAI | 38.5% | Not reported | Closed |
| 55 | Qwen3.8-Flash-NextAlibaba | 38.0% | Not reported | Open |
| 56 | GPT-5.2OpenAI | 37.7% | Not reported | Closed |
| 57 | Kimi K2.6Moonshot AI | 37.5% | Not reported | Open |
| 58 | Grok 4.3xAI | 37.2% | Not reported | Closed |
| 59 | GPT-5.2-CodexOpenAI | 35.7% | Not reported | Closed |
| 60 | MiMo-V2.5-ProXiaomi | 35.7% | Not reported | Closed |
| 61 | Qwen3.7 PlusAlibaba | 35.6% | Not reported | Closed |
| 62 | DeepSeek V4 Pro (High)DeepSeek | 35.2% | Not reported | Open |
| 63 | MiMo-V2.6-FlashXiaomi | 35.1% | Not reported | Open |
| 64 | Kimi K2.7 CodeMoonshot AI | 35.0% | Not reported | Open |
| 65 | Mistral Large 4Mistral | 35.0% | Not reported | Pending |
| 66 | Beam 1.0 501B-A23BReflection AI | 34.7% | Not reported | Open |
| 67 | Apodex 1.1Apodex | 34.1% | Not reported | Closed |
| 68 | Apodex 1.1 MiniApodex | 34.1% | Not reported | Open |
| 69 | Qwen3.8-27BAlibaba | 33.9% | Not reported | Open |
| 70 | Hy3Tencent | 33.5% | Not reported | Open |
| 71 | Hy3 PreviewTencent | 33.5% | Not reported | Open |
| 72 | Inkling-SmallThinking Machines Lab | 33.3% | Not reported | Open |
| 73 | Claude Opus 4.7Anthropic | 33.3% | Not reported | Closed |
| 74 | InklingThinking Machines Lab | 31.9% | Not reported | Open |
| 75 | Qwen 3.6 Max (preview)Alibaba | 30.8% | Not reported | Closed |
| 76 | Kimi K2.5 (Reasoning)Moonshot AI | 30.7% | Not reported | Closed |
| 77 | Kimi K2.5Moonshot AI | 30.7% | Not reported | Open |
| 78 | Nemotron 3.5 Super VL 120B A12BNVIDIA | 30.7% | Not reported | Open |
| 79 | Nemotron 3.5 Super VL 120B A12B NVFP4NVIDIA | 30.7% | Not reported | Open |
| 80 | MiMo-V2-ProXiaomi | 30.4% | Not reported | Closed |
| 81 | GLM-5.1Z.AI | 30.1% | Not reported | Open |
| 82 | Claude Opus 4.5 ThinkingAnthropic | 30.1% | Not reported | Closed |
| 83 | Solar Pro 4Upstage | 29.8% | Not reported | Closed |
| 84 | MiniMax M2.7MiniMax | 29.6% | Not reported | Open |
| 85 | A.X K2SK Telecom | 29.6% | Not reported | Open |
| 86 | GLM-5Z.AI | 29.3% | Not reported | Open |
| 87 | Qwen3.5 397BAlibaba | 29.0% | Not reported | Open |
| 88 | Qwen3.5 397B (Reasoning)Alibaba | 29.0% | Not reported | Open |
| 89 | GPT-5.1OpenAI | 28.5% | Not reported | Closed |
| 90 | GPT-5 (high)OpenAI | 28.5% | Not reported | Closed |
| 91 | Nemotron 3 UltraNVIDIA | 28.4% | Not reported | Open |
| 92 | GPT-5.4 nanoOpenAI | 28.3% | Not reported | Closed |
| 93 | GPT-5.4 miniOpenAI | 28.1% | Not reported | Closed |
| 94 | Qwen3.6 PlusAlibaba | 27.8% | Not reported | Closed |
| 95 | GLM-5-TurboZ.AI | 27.8% | Not reported | Closed |
| 96 | GLM-4.7Z.AI | 27.4% | Not reported | Open |
| 97 | Grok 4xAI | 26.7% | Not reported | Closed |
| 98 | GPT-5.1-CodexOpenAI | 25.7% | Not reported | Closed |
| 99 | GPT-5.1-Codex-MaxOpenAI | 25.7% | Not reported | Closed |
| 100 | GPT-5 (medium)OpenAI | 25.4% | Not reported | Closed |
| 101 | Qwen3.5-122B-A10BAlibaba | 25.2% | Not reported | Open |
| 102 | Nemotron 3.5 Bolt 74B-A9BNVIDIA | 24.7% | Not reported | Open |
| 103 | Step 3.5 FlashStepFun | 24.5% | Not reported | Open |
| 104 | Qwen3.5-27BAlibaba | 23.9% | Not reported | Open |
| 105 | Ling 3.0 FlashInclusionAI | 23.7% | Not reported | Open |
| 106 | Gemma 4 31BGoogle | 23.6% | Not reported | Open |
| 107 | Qwen3.6-27BAlibaba | 23.1% | Not reported | Open |
| 108 | Gemini 2.5 ProGoogle | 22.5% | Not reported | Closed |
| 109 | Qwen3.6-35B-A3BAlibaba | 22.2% | Not reported | Open |
| 110 | MiMo-V2-OmniXiaomi | 22.1% | Not reported | Closed |
| 111 | Muse Glimmer 30BMeta | 22.0% | Not reported | Open |
| 112 | Ling 3.0 Flash VLInclusionAI | 22.0% | Not reported | Open |
| 113 | GPT-5 miniOpenAI | 21.5% | Not reported | Closed |
| 114 | Step 3.7 FlashStepFun | 21.4% | Not reported | Open |
| 115 | Qwen3.5-35B-A3BAlibaba | 21.0% | Not reported | Open |
| 116 | MiniMax M2.5MiniMax | 20.5% | Not reported | Closed |
| 117 | o3OpenAI | 20.1% | Not reported | Closed |
| 118 | GPT-OSS 120BOpenAI | 19.6% | Not reported | Open |
| 119 | Gemma 4 26B A4BGoogle | 19.3% | Not reported | Open |
| 120 | Grok 4.1 Fast (Reasoning)xAI | 19.3% | Not reported | Closed |
| 121 | Claude Opus 4.6Anthropic | 19.1% | Not reported | Closed |
| 122 | Grok 4 Fast (Reasoning)xAI | 19.1% | Not reported | Closed |
| 123 | Gemini 3.5 Flash-LiteGoogle | 18.8% | Not reported | Closed |
| 124 | Quasar 438BMultiverse Computing | 18.7% | Not reported | Closed |
| 125 | K-EXAONE 2.0LG AI Research | 18.6% | Not reported | Open |
| 126 | GLM-5V-TurboZ.AI | 17.1% | Not reported | Closed |
| 127 | Mercury 2Inception | 17.1% | Not reported | Closed |
| 128 | o4-miniOpenAI | 16.5% | Not reported | Closed |
| 129 | DeepSeek-R1DeepSeek | 15.8% | Not reported | Open |
| 130 | Trinity-Large-ThinkingArcee AI | 15.8% | Not reported | Open |
| 131 | Trinity-Large-PreviewArcee AI | 15.8% | Not reported | Open |
| 132 | Gemma 4 12BGoogle | 15.7% | Not reported | Open |
| 133 | Gemini 3 FlashGoogle | 15.0% | Not reported | Closed |
| 134 | DeepSeek V3.1 (Reasoning)DeepSeek | 14.3% | Not reported | Open |
| 135 | K-ExaoneLG AI Research | 13.9% | Not reported | Closed |
| 136 | Mistral Medium 3.5 128BMistral | 13.8% | Not reported | Open |
| 137 | Claude Sonnet 4.6Anthropic | 13.3% | Not reported | Closed |
| 138 | Claude Opus 4.5Anthropic | 13.2% | Not reported | Closed |
| 139 | Claude 4.1 Opus ThinkingAnthropic | 12.5% | Not reported | Closed |
| 140 | Claude 4 Opus ThinkingAnthropic | 12.3% | Not reported | Closed |
| 141 | Command A+Cohere | 12.0% | Not reported | Open |
| 142 | Qwen3 MaxAlibaba | 11.9% | Not reported | Closed |
| 143 | Mercury 2.5Inception | 11.8% | Not reported | Closed |
| 144 | Granite 4.2 30BIBM | 11.2% | Not reported | Open |
| 145 | DeepSeek V3.2DeepSeek | 11.2% | Not reported | Open |
| 146 | North Mini CodeCohere | 11.1% | Not reported | Open |
| 147 | GPT-OSS 20BOpenAI | 11.0% | Not reported | Open |
| 148 | Sarvam 105BSarvam | 11.0% | Not reported | Open |
| 149 | Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA | 10.6% | Not reported | Open |
| 150 | Solar Pro 3Upstage | 10.3% | Not reported | Closed |
| 151 | Qwen3.5-4BAlibaba | 9.9% | Not reported | Open |
| 152 | Mistral Small 4Mistral | 9.9% | Not reported | Open |
| 153 | Magistral MediumMistral | 9.8% | Not reported | Closed |
| 154 | Claude 3.7 Sonnet ThinkingAnthropic | 9.7% | Not reported | Closed |
| 155 | Granite 4.2 8BIBM | 9.7% | Not reported | Open |
| 156 | GPT-5 nanoOpenAI | 9.5% | Not reported | Closed |
| 157 | Ling 3.0 TinyInclusionAI | 9.3% | Not reported | Open |
| 158 | MiniCPM5-2BOpenBMB | 8.9% | Not reported | Open |
| 159 | MiniMax M1 80kMiniMax | 8.9% | Not reported | Closed |
| 160 | MiMo-V2-FlashXiaomi | 8.6% | Not reported | Open |
| 161 | Grok Code Fast 1xAI | 8.0% | Not reported | Closed |
| 162 | o3-miniOpenAI | 7.9% | Not reported | Closed |
| 163 | Magistral SmallMistral | 7.6% | Not reported | Open |
| 164 | GLM-4.7-FlashZ.AI | 7.6% | Not reported | Open |
| 165 | Sarvam 30BSarvam | 7.5% | Not reported | Open |
| 166 | Qwen3-Omni-30B-A3B-ThinkingAlibaba | 7.5% | Not reported | Open |
| 167 | Kimi K2Moonshot AI | 7.4% | Not reported | Closed |
| 168 | Nemotron Ultra 253BNVIDIA | 7.4% | Not reported | Open |
| 169 | QwQ-32BAlibaba | 7.4% | Not reported | Open |
| 170 | o1OpenAI | 7.0% | Not reported | Closed |
| 171 | GLM-4.5-AirZ.AI | 7.0% | Not reported | Closed |
| 172 | LFM2.5-8B-A1BLiquidAI | 6.9% | Not reported | Open |
| 173 | Celeris-1Celeris | 6.8% | Not reported | Closed |
| 174 | DeepSeek V3.1DeepSeek | 6.7% | Not reported | Open |
| 175 | LFM2.5-1.2B-InstructLiquidAI | 6.7% | Not reported | Closed |
| 176 | Granite 4.2 3BIBM | 6.6% | Not reported | Open |
| 177 | Granite-4.0-H-350MIBM | 6.4% | Not reported | Open |
| 178 | Ling 2.6 FlashInclusionAI | 6.3% | Not reported | Open |
| 179 | Claude 4 OpusAnthropic | 6.2% | Not reported | Closed |
| 180 | LFM2.5-2.6BLiquidAI | 6.2% | Not reported | Open |
| 181 | LFM2.5-1.2B-ThinkingLiquidAI | 6.2% | Not reported | Closed |
| 182 | Exaone 4.0 1.2BLG AI Research | 5.7% | Not reported | Open |
| 183 | GLM-4.6Z.AI | 5.5% | Not reported | Open |
| 184 | Granite-4.0-350MIBM | 5.5% | Not reported | Open |
| 185 | Ministral 3 3B (Reasoning)Mistral | 5.4% | Not reported | Open |
| 186 | Ministral 3 3BMistral | 5.4% | Not reported | Open |
| 187 | Grok 4.1 FastxAI | 5.1% | Not reported | Closed |
| 188 | LFM2.5-VL-1.6B-ExtractLiquidAI | 5.1% | Not reported | Open |
| 189 | Exaone 4.0 32BLG AI Research | 5.0% | Not reported | Open |
| 190 | GPT-4.1 miniOpenAI | 5.0% | Not reported | Closed |
| 191 | Granite-4.0-H-1BIBM | 5.0% | Not reported | Open |
| 192 | Phi-4 Multimodal InstructMicrosoft | 5.0% | Not reported | Open |
| 193 | Llama 4 MaverickMeta | 4.9% | Not reported | Open |
| 194 | Nemotron 3 Nano Omni 30B A3BNVIDIA | 4.8% | Not reported | Open |
| 195 | Gemma 4 E2BGoogle | 4.8% | Not reported | Open |
| 196 | DeepSeek V3 0324DeepSeek | 4.7% | Not reported | Open |
| 197 | Gemini 2.5 FlashGoogle | 4.7% | Not reported | Closed |
| 198 | Gemini 1.5 ProGoogle | 4.6% | Not reported | Closed |
| 199 | Nemotron 3 Nano 30BNVIDIA | 4.6% | Not reported | Open |
| 200 | Qwen3-Omni-30B-A3B-InstructAlibaba | 4.6% | Not reported | Open |
| 201 | DeepSeek R1 Distill Qwen 32BDeepSeek | 4.6% | Not reported | Open |
| 202 | Ministral 3 14B (Reasoning)Mistral | 4.6% | Not reported | Open |
| 203 | Ministral 3 14BMistral | 4.6% | Not reported | Open |
| 204 | Gemma 3 27BGoogle | 4.4% | Not reported | Open |
| 205 | Claude 4 SonnetAnthropic | 4.3% | Not reported | Closed |
| 206 | Llama 3.1 8BMeta | 4.3% | Not reported | Open |
| 207 | Gemini 2.0 FlashGoogle | 4.3% | Not reported | Closed |
| 208 | Mistral Small 3.1Mistral | 4.3% | Not reported | Open |
| 209 | Ministral 3 8B (Reasoning)Mistral | 4.3% | Not reported | Open |
| 210 | Ministral 3 8BMistral | 4.3% | Not reported | Open |
| 211 | GPT-4.1OpenAI | 4.2% | Not reported | Closed |
| 212 | Claude 3.7 SonnetAnthropic | 4.2% | Not reported | Closed |
| 213 | Mistral Large 3Mistral | 4.2% | Not reported | Closed |
| 214 | GPT-4o miniOpenAI | 4.2% | Not reported | Closed |
| 215 | Gemini 1.0 ProGoogle | 4.2% | Not reported | Closed |
| 216 | LFM2-24B-A2BLiquidAI | 4.2% | Not reported | Closed |
| 217 | Mistral Medium 3Mistral | 4.1% | Not reported | Closed |
| 218 | Claude 3 HaikuAnthropic | 4.1% | Not reported | Closed |
| 219 | Llama 3.1 405BMeta | 4.0% | Not reported | Open |
| 220 | Gemma 4 E4BGoogle | 3.8% | Not reported | Open |
| 221 | Phi-4Microsoft | 3.8% | Not reported | Open |
| 222 | Llama 4 ScoutMeta | 3.8% | Not reported | Open |
| 223 | Phi-4-reasoningMicrosoft | 3.8% | Not reported | Open |
| 224 | GPT-4.1 nanoOpenAI | 3.8% | Not reported | Closed |
| 225 | Devstral SmallMistral | 3.8% | Not reported | Open |
| 226 | Mistral Small 4 (Reasoning)Mistral | 3.8% | Not reported | Open |
| 227 | Gemini 2.5 Flash-LiteGoogle | 3.7% | Not reported | Closed |
| 228 | Solar Pro 2Upstage | 3.7% | Not reported | Closed |
| 229 | Llama 3.3 70BMeta | 3.6% | Not reported | Open |
| 230 | Llama 3.1 70BMeta | 3.6% | Not reported | Open |
| 231 | o1-miniOpenAI | 3.6% | Not reported | Closed |
| 232 | Claude 3.5 HaikuAnthropic | 3.6% | Not reported | Closed |
| 233 | Ultravox v0.6 Llama 3.3 70BFixie AI | 3.6% | Not reported | Open |
| 234 | Qwen2.5 Coder 32B InstructAlibaba | 3.5% | Not reported | Open |
| 235 | Mistral Large 2Mistral | 3.3% | Not reported | Closed |
| 236 | Nova ProAmazon | 3.2% | Not reported | Closed |
| 237 | Gemini 1.5 FlashGoogle | 3.2% | Not reported | Closed |
| 238 | GPT-4 TurboOpenAI | 3.1% | Not reported | Closed |
| 239 | DeepSeek V3DeepSeek | 2.9% | Not reported | Open |
| 240 | Claude 3 OpusAnthropic | 2.8% | Not reported | Closed |
| 241 | GPT-4oOpenAI | 2.4% | Not reported | Closed |
According to BenchLM.ai, Claude Opus 5.5 leads the AA-HLE benchmark with a score of 61.4%, followed by Gemini 4 Carbon (60.4%) and Claude Fable 5.1 (59.1%). The top models are clustered within 2.3 points, suggesting this benchmark is nearing saturation for frontier models.
241 models have been evaluated on AA-HLE. The benchmark falls in the Knowledge category. BenchAlign v5.8 gives AA-HLE 11% of the Knowledge reference weight, so it moves the Knowledge leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.
About AA-HLE
Year
2026
Tasks
Expert-level questions
Format
Accuracy
Difficulty
Frontier expert reasoning
BenchLM stores the Artificial Analysis HLE result separately from the core HLE lane because Artificial Analysis controls the evaluation configuration.
Freshness and provenance
Version
AA-HLE 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
Current
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does AA-HLE measure?
An Artificial Analysis Humanity's Last Exam score.
Which model scores highest on AA-HLE?
Claude Opus 5.5 by Anthropic currently leads with a score of 61.4% on AA-HLE.
How many models are evaluated on AA-HLE?
241 AI models have published results on AA-HLE in the BenchLM catalog.
Last updated: October 28, 2026 · BenchLM version AA-HLE 2026
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.