Skip to main content
idapt
AccueilModèles IATarifs
Sign inStart free
All terms

Learn

What is the AIME benchmark?

Last reviewed July 16, 2026

The AIME (American Invitational Mathematics Examination) is a high-school competition exam used to benchmark AI mathematical reasoning: 30 problems per year (two 15-problem papers), each with an integer answer from 0 to 999, demanding multi-step reasoning rather than recall. Because a benchmark year has only 30 problems, each one is worth over 3 percentage points, so small score gaps between models are noise.

Why AIME became an AI benchmark

It is genuinely hard multi-step math with objectively checkable integer answers, refreshed every year, which limits contamination for the newest edition. Reasoning models made it the headline math test: scores that sat in single digits before reasoning-token training now reach most of the paper, so AIME is where that capability jump is most visible.

The small-N caveat

Thirty problems is a tiny sample: one lucky or unlucky problem moves a score by 3.3 points, and run-to-run variance (sampling, retries) adds more. Treat differences under several points as ties, prefer results averaged over multiple runs, and check which year's exam a number refers to before comparing.

Top models on AIME

Full leaderboard
  1. 1GPT 5.5100.0%
  2. 2GPT 5.5 Pro100.0%
  3. 3Claude Fable 599.7%
  4. 4Claude Opus 4.898.3%
  5. 5Claude Opus 4.797.8%
  6. 6Grok 4.597.8%
  7. 7DeepSeek V4 Pro96.7%
  8. 8Kimi K2.7 Code96.4%
  9. 9Kimi K2.696.1%
  10. 10GPT 5.296.1%

Peak scores from the compiled benchmark data; reasoning models use their highest effort tier.

In idapt

idapt's math rankings include an AIME lane with the exact score per model, and reasoning models show per-effort-tier numbers on their pages: effort visibly moves AIME results.

AIME leaderboardMath rankings

Frequently asked

Related terms

MATH-500FrontierMathReasoning effort

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Code Execution

Sandboxed Python, Node, and shell: agents and you, side-by-side.

See it in practice

The fastest way to understand it is to use it. Start free, no credit card.

Try idapt free
  • Accueil
  • Tarifs
  • Modèles IA
  • Modèles d'image
  • Modèles de voix
  • Modèles vidéo
  • Classements
  • New models
  • Model status
  • Multi-Model Chat
  • Voice Mode
  • Voice HUD
  • Web Search
  • Image Generation
  • Video Generation
  • Audio Generation
  • Transcription
  • Drive
  • Secrets
  • Sharing
  • Workspaces
  • Tasks
  • Memory
  • Agents
  • Subagents
  • Automations
  • Skills
  • Code Execution
  • Computers
  • Computer Use
  • Cloud Computers
  • Local AI
  • AI Gateway
  • API & SDK
  • CLI
  • MCP
  • Tunnels
  • All features →
  • Blog
  • Cas d'usage
  • Comparisons
  • Best of
  • Learn
  • Free tools
  • Prompts
  • Templates
  • Changelog
  • Centre d'aide
  • FAQ
  • Confidentialité
  • Comparer tous les modèles
  • Support
  • Développeurs
  • Quickstarts
  • Référence API
  • Tarifs API
  • CLI
  • MCP
  • Downloads
  • Desktop
  • Badges and embeds
© idapt[email protected]ConditionsPolitique de confidentialitéMentions légalesSignaler un contenu
X (Twitter)