evals.report
BenchmarksLabsCompareRun guidesIn the wild
BenchmarksReasoning

AIME 2026

Accuracy of LLMs on the 30 problems of the 2026 American Invitational Mathematics Examination (AIME I and II), a contamination-free competition-math benchmark requiring integer answers (0-999), evaluated live by MathArena.

ReasoningaccuracyHigher is better
ScoresAboutRun this benchmark

What is AIME 2026?

Accuracy of LLMs on the 30 problems of the 2026 American Invitational Mathematics Examination (AIME I and II), a contamination-free competition-math benchmark requiring integer answers (0-999), evaluated live by MathArena. evals.report tracks reported AIME 2026 scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported AIME 2026 score: Claude Opus 4.8 — 100.00% (accuracy).

ModelLabScore↓Source modelStatusDate
Claude Opus 4.8Anthropic100.00%—OfficialMay 28, 2026Details
GLM-5.2OpenZ.ai99.2%GLM-5.2VerifiedJun 16, 2026Details
GPT-5.4OpenAI99.17%—OfficialMar 5, 2026Details
GPT-5.2OpenAI98.33%—OfficialDec 11, 2025Details
Gemini 3.1 Pro PreviewGoogle DeepMind98.33%—OfficialFeb 19, 2026Details
GPT-5.5OpenAI97.50%—OfficialApr 23, 2026Details
InklingOpenThinking Machines Lab97.1%InklingVerifiedJul 15, 2026Details
Claude Opus 4.6Anthropic96.67%—OfficialFeb 5, 2026Details
DeepSeek V4 FlashOpenDeepSeek95.83%—OfficialApr 24, 2026Details
Gemini 3 FlashGoogle DeepMind95.83%—OfficialDec 17, 2025Details
Kimi K2.5OpenMoonshot AI95.83%—OfficialJan 27, 2026Details
GLM-5OpenZ.ai95.83%—OfficialFeb 11, 2026Details
DeepSeek V4 ProOpenDeepSeek95.83%—OfficialApr 24, 2026Details
GLM-5.1OpenZ.ai95.83%—OfficialApr 7, 2026Details
Kimi K2.6OpenMoonshot AI95.83%—OfficialApr 20, 2026Details
Claude Opus 4.7Anthropic95.83%—OfficialApr 16, 2026Details
Gemini 3.5 FlashGoogle DeepMind95.00%—OfficialMay 19, 2026Details
Muse Glimmer 30BOpenMeta94.7%Muse Glimmer-30B (High Reasoning)VerifiedAug 10, 2026Details
Grok 4.1 fast reasoningxAI94.17%—OfficialNov 19, 2025Details
DeepSeek V3.2OpenDeepSeek94.17%—OfficialDec 1, 2025Details
Qwen3.5-397B-A17BOpenAlibaba / Qwen93.33%—OfficialFeb 16, 2026Details
Gemini 3 ProGoogle DeepMind91.67%—OfficialNov 18, 2025Details
NVIDIA Nemotron 3 Super 120B-A12BOpenNVIDIA90.00%—OfficialMar 10, 2026Details

Each row reports the model’s accuracy on AIME 2026. Click a row for the full run context.