Scores over time

Best recorded score per model; higher is better. The line follows the running best. Evaluation contexts can differ.

Humanity's Last Exam Accuracy scores over timeOne point per model's best recorded score, dated by reported date or model release date. A line shows the running best. The data table below contains every plotted point.64.410.62025-11-182026-10-07Gemini 3 Pro: 37.5% · Reported 2025-11-18Claude Opus 4.5: 30.8% · Reported 2025-11-24NVIDIA Nemotron 3 Nano 30B-A3B: 10.6% · Reported 2025-12-15Gemini 3 Flash: 33.7% · Reported 2025-12-17Qwen3-Max-Thinking: 30.2% · Reported 2026-01-25Kimi K2.5: 30.1% · Reported 2026-01-27Claude Opus 4.6: 40.0% · Reported 2026-02-05Claude Sonnet 4.6: 33.2% · Reported 2026-02-17Qwen3.5-397B-A17B: 28.7% · Reported 2026-02-17Gemini 3.1 Pro: 44.4% · Reported 2026-02-19GPT-5.4: 39.8% · Reported 2026-03-05NVIDIA Nemotron 3 Super 120B-A12B: 18.3% · Reported 2026-03-11Muse Spark: 50.2% · Reported 2026-04-08Claude Opus 4.7: 46.9% · Reported 2026-04-16Kimi K2.6: 34.7% · Reported 2026-04-20GPT-5.5: 41.4% · Reported 2026-04-23DeepSeek-V4-Flash: 34.8% · Reported 2026-04-24DeepSeek-V4-Pro: 37.7% · Reported 2026-04-24Gemini 3.5 Flash: 40.2% · Reported 2026-05-19Claude Opus 4.8: 49.8% · Reported 2026-05-28NVIDIA Nemotron 3 Ultra 550B-A55B: 26.7% · Reported 2026-06-04Claude Sonnet 5: 43.2% · Reported 2026-06-30Kimi K3: 43.5% · Reported 2026-07-16Claude Opus 5: 56.4% · Reported 2026-07-24Qwen3.8-Max: 43.6% · Reported 2026-08-02Claude Fable 5.1: 60.9% · Reported 2026-09-01Claude Opus 5.5: 64.4% · Reported 2026-09-22Claude Sonnet 5.5: 56.9% · Reported 2026-09-28Claude Haiku 5.5: 45.9% · Reported 2026-10-07
Data table
Scores plotted over time
ModelDateScore
Gemini 3 Pro (Reported)37.5%
Claude Opus 4.5 (Reported)30.8%
NVIDIA Nemotron 3 Nano 30B-A3B (Reported)10.6%
Gemini 3 Flash (Reported)33.7%
Qwen3-Max-Thinking (Reported)30.2%
Kimi K2.5 (Reported)30.1%
Claude Opus 4.6 (Reported)40.0%
Claude Sonnet 4.6 (Reported)33.2%
Qwen3.5-397B-A17B (Reported)28.7%
Gemini 3.1 Pro (Reported)44.4%
GPT-5.4 (Reported)39.8%
NVIDIA Nemotron 3 Super 120B-A12B (Reported)18.3%
Muse Spark (Reported)50.2%
Claude Opus 4.7 (Reported)46.9%
Kimi K2.6 (Reported)34.7%
GPT-5.5 (Reported)41.4%
DeepSeek-V4-Flash (Reported)34.8%
DeepSeek-V4-Pro (Reported)37.7%
Gemini 3.5 Flash (Reported)40.2%
Claude Opus 4.8 (Reported)49.8%
NVIDIA Nemotron 3 Ultra 550B-A55B (Reported)26.7%
Claude Sonnet 5 (Reported)43.2%
Kimi K3 (Reported)43.5%
Claude Opus 5 (Reported)56.4%
Qwen3.8-Max (Reported)43.6%
Claude Fable 5.1 (Reported)60.9%
Claude Opus 5.5 (Reported)64.4%
Claude Sonnet 5.5 (Reported)56.9%
Claude Haiku 5.5 (Reported)45.9%

Results

46 results

Humanity's Last Exam Full set — No tools results
ProviderModelScoreSourceRegistry No.
AnthropicClaude Haiku 5.5 (max)45.9%anthropic.com for Claude Haiku 5.5 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Haiku 5.5 on Humanity's Last Exam Full set — No tools20018
AnthropicClaude Sonnet 5.5 (medium)43.2%anthropic.com for Claude Sonnet 5.5 (medium) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5.5 on Humanity's Last Exam Full set — No tools20016
AnthropicClaude Sonnet 5.5 (max)56.9%anthropic.com for Claude Sonnet 5.5 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5.5 on Humanity's Last Exam Full set — No tools20016
AnthropicClaude Sonnet 5.5 (low)41.4%anthropic.com for Claude Sonnet 5.5 (low) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5.5 on Humanity's Last Exam Full set — No tools20016
AnthropicClaude Sonnet 5.5 (high)47.9%anthropic.com for Claude Sonnet 5.5 (high) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5.5 on Humanity's Last Exam Full set — No tools20016
AnthropicClaude Sonnet 5.5 (xhigh)53.0%anthropic.com for Claude Sonnet 5.5 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5.5 on Humanity's Last Exam Full set — No tools20016
AnthropicClaude Opus 5.5 (medium)59.0%www-cdn.anthropic.com for Claude Opus 5.5 (medium) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5.5 on Humanity's Last Exam Full set — No tools20015
AnthropicClaude Opus 5.5 (max)64.4%www-cdn.anthropic.com for Claude Opus 5.5 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5.5 on Humanity's Last Exam Full set — No tools20015
AnthropicClaude Opus 5.5 (high)59.6%www-cdn.anthropic.com for Claude Opus 5.5 (high) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5.5 on Humanity's Last Exam Full set — No tools20015
AnthropicClaude Opus 5.5 (low)52.9%www-cdn.anthropic.com for Claude Opus 5.5 (low) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5.5 on Humanity's Last Exam Full set — No tools20015
AnthropicClaude Opus 5.5 (xhigh)62.8%www-cdn.anthropic.com for Claude Opus 5.5 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5.5 on Humanity's Last Exam Full set — No tools20015
AnthropicClaude Fable 5.1 (high)58.0%anthropic.com for Claude Fable 5.1 (high) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAnthropicSelf-reportedRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — No tools20004
AnthropicClaude Fable 5.1 (medium)55.9%anthropic.com for Claude Fable 5.1 (medium) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAnthropicSelf-reportedRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — No tools20004
AnthropicClaude Fable 5.1 (low)53.2%anthropic.com for Claude Fable 5.1 (low) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAnthropicSelf-reportedRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — No tools20004
AnthropicClaude Fable 5.1 (no tools)60.9%anthropic.com for Claude Fable 5.1 (no tools) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAnthropicSelf-reportedRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — No tools20004
AnthropicClaude Fable 5.1 (xhigh)60.4%anthropic.com for Claude Fable 5.1 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAnthropicSelf-reportedRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — No tools20004
Alibaba (Tongyi)Qwen3.8-Max43.6%huggingface.co for Qwen3.8-Max on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardQwenSelf-reportedQwen TeamRecord for Qwen3.8-Max on Humanity's Last Exam Full set — No tools130008
AnthropicClaude Opus 5 (low)47.8%www-cdn.anthropic.com for Claude Opus 5 (low) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5 on Humanity's Last Exam Full set — No tools20014
AnthropicClaude Opus 5 (high)56.0%www-cdn.anthropic.com for Claude Opus 5 (high) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5 on Humanity's Last Exam Full set — No tools20014
AnthropicClaude Opus 5 (xhigh)56.4%www-cdn.anthropic.com for Claude Opus 5 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5 on Humanity's Last Exam Full set — No tools20014
AnthropicClaude Opus 5 (medium)54.2%www-cdn.anthropic.com for Claude Opus 5 (medium) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5 on Humanity's Last Exam Full set — No tools20014
AnthropicClaude Opus 5 (max)56.3%www-cdn.anthropic.com for Claude Opus 5 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 5 on Humanity's Last Exam Full set — No tools20014
Moonshot AIKimi K3 (max)43.5%huggingface.co for Kimi K3 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardMoonshot AISelf-reportedRecord for Kimi K3 on Humanity's Last Exam Full set — No tools120003
AnthropicClaude Sonnet 5 (max)43.2%www-cdn.anthropic.com for Claude Sonnet 5 (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 5 on Humanity's Last Exam Full set — No tools20013
NVIDIANVIDIA Nemotron 3 Ultra 550B-A55B (no-tools)26.7%huggingface.co for NVIDIA Nemotron 3 Ultra 550B-A55B (no-tools) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardNVIDIASelf-reportedRecord for NVIDIA Nemotron 3 Ultra 550B-A55B on Humanity's Last Exam Full set — No tools60003
AnthropicClaude Opus 4.8 (adaptive thinking, max)49.8%www-cdn.anthropic.com for Claude Opus 4.8 (adaptive thinking, max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 4.8 on Humanity's Last Exam Full set — No tools20011
Google (DeepMind)Gemini 3.5 Flash40.2%deepmind.google for Gemini 3.5 Flash on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardGoogle DeepMindSelf-reportedRecord for Gemini 3.5 Flash on Humanity's Last Exam Full set — No tools30006
DeepSeek AIDeepSeek-V4-Flash (max)34.8%huggingface.co for DeepSeek-V4-Flash (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardDeepSeekSelf-reportedDeepSeek AIRecord for DeepSeek-V4-Flash on Humanity's Last Exam Full set — No tools110001
DeepSeek AIDeepSeek-V4-Pro (max)37.7%huggingface.co for DeepSeek-V4-Pro (max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardDeepSeekSelf-reportedDeepSeek AIRecord for DeepSeek-V4-Pro on Humanity's Last Exam Full set — No tools110002
OpenAIGPT-5.5 (xhigh)41.4%openai.com for GPT-5.5 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postOpenAISelf-reportedRecord for GPT-5.5 on Humanity's Last Exam Full set — No tools10008
Moonshot AIKimi K2.6 (thinking)34.7%huggingface.co for Kimi K2.6 (thinking) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardMoonshot AISelf-reportedRecord for Kimi K2.6 on Humanity's Last Exam Full set — No tools120002
AnthropicClaude Opus 4.7 (adaptive thinking, max)46.9%www-cdn.anthropic.com for Claude Opus 4.7 (adaptive thinking, max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 4.7 on Humanity's Last Exam Full set — No tools20010
Meta AI (originally Facebook AI Research)Muse Spark (contemplating)50.2%ai.meta.com for Muse Spark (contemplating) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postMetaSelf-reportedRecord for Muse Spark on Humanity's Last Exam Full set — No tools80001
Meta AI (originally Facebook AI Research)Muse Spark (thinking)42.8%ai.meta.com for Muse Spark (thinking) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Evaluation methodologyMetaSelf-reportedRecord for Muse Spark on Humanity's Last Exam Full set — No tools80001
NVIDIANVIDIA Nemotron 3 Super 120B-A12B (no-tools)18.3%huggingface.co for NVIDIA Nemotron 3 Super 120B-A12B (no-tools) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardNVIDIASelf-reportedRecord for NVIDIA Nemotron 3 Super 120B-A12B on Humanity's Last Exam Full set — No tools60002
OpenAIGPT-5.4 (xhigh)39.8%openai.com for GPT-5.4 (xhigh) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postOpenAISelf-reportedRecord for GPT-5.4 on Humanity's Last Exam Full set — No tools10007
Google (DeepMind)Gemini 3.1 Pro (High)44.4%deepmind.google for Gemini 3.1 Pro (High) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardGoogle DeepMindSelf-reportedRecord for Gemini 3.1 Pro on Humanity's Last Exam Full set — No tools30005
Alibaba (Tongyi)Qwen3.5-397B-A17B28.7%alibabacloud.com for Qwen3.5-397B-A17B on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAlibaba Cloud CommunitySelf-reportedQwen TeamRecord for Qwen3.5-397B-A17B on Humanity's Last Exam Full set — No tools130003
AnthropicClaude Sonnet 4.6 (adaptive thinking, max)33.2%www-cdn.anthropic.com for Claude Sonnet 4.6 (adaptive thinking, max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Sonnet 4.6 on Humanity's Last Exam Full set — No tools20009
AnthropicClaude Opus 4.6 (adaptive thinking, max)40.0%www-cdn.anthropic.com for Claude Opus 4.6 (adaptive thinking, max) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 4.6 on Humanity's Last Exam Full set — No tools20008
Moonshot AIKimi K2.5 (thinking)30.1%huggingface.co for Kimi K2.5 (thinking) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardMoonshot AISelf-reportedRecord for Kimi K2.5 on Humanity's Last Exam Full set — No tools120001
Alibaba (Tongyi)Qwen3-Max-Thinking30.2%alibabacloud.com for Qwen3-Max-Thinking on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postAlibaba Cloud CommunitySelf-reportedQwen TeamRecord for Qwen3-Max-Thinking on Humanity's Last Exam Full set — No tools130002
Google (DeepMind)Gemini 3 Flash33.7%blog.google for Gemini 3 Flash on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postGoogleSelf-reportedGoogle DeepMindRecord for Gemini 3 Flash on Humanity's Last Exam Full set — No tools30009
NVIDIANVIDIA Nemotron 3 Nano 30B-A3B (no-tools)10.6%huggingface.co for NVIDIA Nemotron 3 Nano 30B-A3B (no-tools) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Model cardNVIDIASelf-reportedRecord for NVIDIA Nemotron 3 Nano 30B-A3B on Humanity's Last Exam Full set — No tools60001
AnthropicClaude Opus 4.5 (extended thinking)30.8%assets.anthropic.com for Claude Opus 4.5 (extended thinking) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported System cardAnthropicSelf-reportedRecord for Claude Opus 4.5 on Humanity's Last Exam Full set — No tools20007
Google (DeepMind)Gemini 3 Pro (high)37.5%blog.google for Gemini 3 Pro (high) on Humanity's Last Exam Full set — No tools (opens in a new tab)Humanity's Last Exam AccuracyReported Launch postGoogleSelf-reportedGoogle DeepMindRecord for Gemini 3 Pro on Humanity's Last Exam Full set — No tools30008