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.54.85.42026-10-032026-10-03OpenAI o4-mini: 18.1% · Reported 2026-10-03GPT-5.2: 27.8% · Reported 2026-10-03GPT-4.5 Preview: 5.4% · Reported 2026-10-03Claude Sonnet 4.5: 13.7% · Reported 2026-10-03Gemini 2.5 Pro: 18.2% · Reported 2026-10-03Gemini 3 Pro: 37.5% · Reported 2026-10-03GPT-5 mini: 19.4% · Reported 2026-10-03GLM-4.5-Air: 8.1% · Reported 2026-10-03Claude Fable 5.1: 46.5% · Reported 2026-10-03Muse Spark: 40.6% · Reported 2026-10-03Claude Opus 4.7: 36.2% · Reported 2026-10-03Claude Opus 4.1: 11.5% · Reported 2026-10-03Claude Sonnet 4: 7.8% · Reported 2026-10-03GPT-5.4: 36.2% · Reported 2026-10-03Claude Opus 4.6: 34.4% · Reported 2026-10-03GPT-5: 25.3% · Reported 2026-10-03Gemini 3.8 Flash: 44.5% · Reported 2026-10-03Claude 3.7 Sonnet: 8.0% · Reported 2026-10-03GPT-4.1: 5.4% · Reported 2026-10-03OpenAI o3: 20.3% · Reported 2026-10-03Claude Opus 4.5: 25.2% · Reported 2026-10-03Gemini 2.5 Flash: 12.1% · Reported 2026-10-03Kimi K2.5: 24.4% · Reported 2026-10-03Claude Opus 4: 10.7% · Reported 2026-10-03Gemini 3.1 Pro: 46.4% · Reported 2026-10-03GPT-6 Astra: 54.8% · Reported 2026-10-03GLM-4.5: 8.3% · Reported 2026-10-03
Data table
Scores plotted over time
ModelDateScore
OpenAI o4-mini (Reported)18.1%
GPT-5.2 (Reported)27.8%
GPT-4.5 Preview (Reported)5.4%
Claude Sonnet 4.5 (Reported)13.7%
Gemini 2.5 Pro (Reported)18.2%
Gemini 3 Pro (Reported)37.5%
GPT-5 mini (Reported)19.4%
GLM-4.5-Air (Reported)8.1%
Claude Fable 5.1 (Reported)46.5%
Muse Spark (Reported)40.6%
Claude Opus 4.7 (Reported)36.2%
Claude Opus 4.1 (Reported)11.5%
Claude Sonnet 4 (Reported)7.8%
GPT-5.4 (Reported)36.2%
Claude Opus 4.6 (Reported)34.4%
GPT-5 (Reported)25.3%
Gemini 3.8 Flash (Reported)44.5%
Claude 3.7 Sonnet (Reported)8.0%
GPT-4.1 (Reported)5.4%
OpenAI o3 (Reported)20.3%
Claude Opus 4.5 (Reported)25.2%
Gemini 2.5 Flash (Reported)12.1%
Kimi K2.5 (Reported)24.4%
Claude Opus 4 (Reported)10.7%
Gemini 3.1 Pro (Reported)46.4%
GPT-6 Astra (Reported)54.8%
GLM-4.5 (Reported)8.3%

Results

35 results

Humanity's Last Exam Full set — Scale AI evaluation results
ProviderModelScoreSourceRegistry No.
OpenAIGPT-4.5 Preview5.4%labs.scale.com for GPT-4.5 Preview on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-4.5 Preview on Humanity's Last Exam Full set — Scale AI evaluation10001
OpenAIGPT-4.15.4%labs.scale.com for GPT-4.1 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-4.1 on Humanity's Last Exam Full set — Scale AI evaluation10002
OpenAIOpenAI o3 (medium)19.2%labs.scale.com for OpenAI o3 (medium) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for OpenAI o3 on Humanity's Last Exam Full set — Scale AI evaluation10003
OpenAIOpenAI o3 (high)20.3%labs.scale.com for OpenAI o3 (high) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for OpenAI o3 on Humanity's Last Exam Full set — Scale AI evaluation10003
OpenAIOpenAI o4-mini (high)18.1%labs.scale.com for OpenAI o4-mini (high) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for OpenAI o4-mini on Humanity's Last Exam Full set — Scale AI evaluation10004
OpenAIOpenAI o4-mini (medium)14.3%labs.scale.com for OpenAI o4-mini (medium) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for OpenAI o4-mini on Humanity's Last Exam Full set — Scale AI evaluation10004
OpenAIGPT-6 Astra54.8%labs.scale.com for GPT-6 Astra on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-6 Astra on Humanity's Last Exam Full set — Scale AI evaluation10005
OpenAIGPT-5.4 (xhigh thinking)36.2%labs.scale.com for GPT-5.4 (xhigh thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-5.4 on Humanity's Last Exam Full set — Scale AI evaluation10007
OpenAIGPT-5.227.8%labs.scale.com for GPT-5.2 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-5.2 on Humanity's Last Exam Full set — Scale AI evaluation10012
OpenAIGPT-525.3%labs.scale.com for GPT-5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-5 on Humanity's Last Exam Full set — Scale AI evaluation10018
OpenAIGPT-5 mini19.4%labs.scale.com for GPT-5 mini on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GPT-5 mini on Humanity's Last Exam Full set — Scale AI evaluation10019
Moonshot AIKimi K2.524.4%labs.scale.com for Kimi K2.5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Kimi K2.5 on Humanity's Last Exam Full set — Scale AI evaluation120001
Z.aiGLM-4.58.3%labs.scale.com for GLM-4.5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GLM-4.5 on Humanity's Last Exam Full set — Scale AI evaluation150007
Z.aiGLM-4.5-Air8.1%labs.scale.com for GLM-4.5-Air on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for GLM-4.5-Air on Humanity's Last Exam Full set — Scale AI evaluation150008
AnthropicClaude 3.7 Sonnet (Thinking)8.0%labs.scale.com for Claude 3.7 Sonnet (Thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude 3.7 Sonnet on Humanity's Last Exam Full set — Scale AI evaluation20001
AnthropicClaude Opus 46.7%labs.scale.com for Claude Opus 4 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4 on Humanity's Last Exam Full set — Scale AI evaluation20002
AnthropicClaude Opus 4 (Thinking)10.7%labs.scale.com for Claude Opus 4 (Thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4 on Humanity's Last Exam Full set — Scale AI evaluation20002
AnthropicClaude Sonnet 4 (Thinking)7.8%labs.scale.com for Claude Sonnet 4 (Thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Sonnet 4 on Humanity's Last Exam Full set — Scale AI evaluation20003
AnthropicClaude Sonnet 45.5%labs.scale.com for Claude Sonnet 4 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Sonnet 4 on Humanity's Last Exam Full set — Scale AI evaluation20003
AnthropicClaude Fable 5.1 (xhigh)46.5%labs.scale.com for Claude Fable 5.1 (xhigh) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Fable 5.1 on Humanity's Last Exam Full set — Scale AI evaluation20004
AnthropicClaude Sonnet 4.5 (thinking)13.7%labs.scale.com for Claude Sonnet 4.5 (thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation20005
AnthropicClaude Sonnet 4.57.5%labs.scale.com for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation20005
AnthropicClaude Opus 4.514.2%labs.scale.com for Claude Opus 4.5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.5 on Humanity's Last Exam Full set — Scale AI evaluation20007
AnthropicClaude Opus 4.5 (thinking)25.2%labs.scale.com for Claude Opus 4.5 (thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.5 on Humanity's Last Exam Full set — Scale AI evaluation20007
AnthropicClaude Opus 4.6 (Non-Thinking)19.0%labs.scale.com for Claude Opus 4.6 (Non-Thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.6 on Humanity's Last Exam Full set — Scale AI evaluation20008
AnthropicClaude Opus 4.6 (thinking-max)34.4%labs.scale.com for Claude Opus 4.6 (thinking-max) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.6 on Humanity's Last Exam Full set — Scale AI evaluation20008
AnthropicClaude Opus 4.736.2%labs.scale.com for Claude Opus 4.7 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.7 on Humanity's Last Exam Full set — Scale AI evaluation20010
AnthropicClaude Opus 4.17.9%labs.scale.com for Claude Opus 4.1 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.1 on Humanity's Last Exam Full set — Scale AI evaluation20017
AnthropicClaude Opus 4.1 (thinking)11.5%labs.scale.com for Claude Opus 4.1 (thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Claude Opus 4.1 on Humanity's Last Exam Full set — Scale AI evaluation20017
Google (DeepMind)Gemini 2.5 Pro18.2%labs.scale.com for Gemini 2.5 Pro on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Gemini 2.5 Pro on Humanity's Last Exam Full set — Scale AI evaluation30002
Google (DeepMind)Gemini 2.5 Flash12.1%labs.scale.com for Gemini 2.5 Flash on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Gemini 2.5 Flash on Humanity's Last Exam Full set — Scale AI evaluation30003
Google (DeepMind)Gemini 3.8 Flash44.5%labs.scale.com for Gemini 3.8 Flash on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Gemini 3.8 Flash on Humanity's Last Exam Full set — Scale AI evaluation30004
Google (DeepMind)Gemini 3.1 Pro (thinking high)46.4%labs.scale.com for Gemini 3.1 Pro (thinking high) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Gemini 3.1 Pro on Humanity's Last Exam Full set — Scale AI evaluation30005
Google (DeepMind)Gemini 3 Pro37.5%labs.scale.com for Gemini 3 Pro on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Gemini 3 Pro on Humanity's Last Exam Full set — Scale AI evaluation30008
Meta AI (originally Facebook AI Research)Muse Spark40.6%labs.scale.com for Muse Spark on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab)Archived copy (opens in a new tab)Humanity's Last Exam AccuracyReported Benchmark leaderboardScale AIIndependentRecord for Muse Spark on Humanity's Last Exam Full set — Scale AI evaluation80001