AI Decathlon - Standard-Judged Tasks: leaderboard
AI Decathlon Timed Challenge (Standard-Judged Tasks) evaluating coding agents on 60-100 minute engineering, math, spatial, and interactive design tasks.
Metric: Program-Verified Score (0-100 across 7 standard tasks). Source: octathlon.wendystars.com. Saturation forecast: Around February 2027. 25 models tracked. Updated 22 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6.2 Astra | xHigh | 88.9 | #4 |
| 2 | GPT-6.2 Sol | xHigh | 70.2 | #10 |
| 3 | GPT-6 Astra | xHigh | 67.2 | #11 |
| 4 | GPT-6.1 Sol | xHigh | 65.4 | #15 |
| 5 | Gemini 4 Neon | xHigh | 65.1 | #3 |
| 6 | Claude Fable 5.2 | xHigh | 57.2 | #2 |
| 7 | Gemini 4 Carbon | High | 48.5 | #7 |
| 8 | Claude Opus 5.5 | xHigh | 45.9 | #9 |
| 9 | Claude Sonnet 5.5 | xHigh | 41 | #24 |
| 10 | Gemini 4 Hydrogen | xHigh | 40.6 | #16 |
| 11 | GPT-6.1 Terra | xHigh | 36.1 | #25 |
| 12 | GPT-5.6 Sol | xHigh | 24.4 | #36 |
| 13 | Claude Haiku 5.5 | xHigh | 19.9 | #75 |
| 14 | Kimi K3.1 Pro | Max | 17.4 | #35 |
| 15 | MiniMax-M3.1-Flash | Max | 15.1 | #119 |
Interactive version: theaggregate.ai/benchmark?slug=ai-decathlon-standard-judged-tasks · How It Works · Data refreshed daily, snapshot 2026-11-22.