Updated daily at 05:00 UTC
Ranking
Best AI models for Coding & software engineering
Writing, fixing and running code in real repositories and terminals, judged by hidden tests.
Ranked models
Based on 5 benchmarks
| Rank | Model | Company | Score | Coverage | Released |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | Anthropic | 62.4 | 3 of 5 | 9 Jun 2026 |
| 2 | GPT-6 Astra | OpenAI | 58.0 | 3 of 5 | 3 Sep 2026 |
| 3 | Claude Opus 4.8 | Anthropic | 50.8 | 3 of 5 | 28 May 2026 |
| 4 | GPT-5.6 Sol | OpenAI | 46.7 | 3 of 5 | 9 Jul 2026 |
| 5 | Claude Sonnet 5 | Anthropic | 44.6 | 3 of 5 | 30 Jun 2026 |
| 6 | GPT-5.5 | OpenAI | 40.0 | 4 of 5 | 23 Apr 2026 |
| 7 | Claude Opus 4.7 | Anthropic | 37.9 | 3 of 5 | 16 Apr 2026 |
| 8 | GPT-5.4 | OpenAI | 32.9 | 3 of 5 | 5 Mar 2026 |
| 9 | Gemini 3.1 Pro | Google DeepMind | 14.4 | 3 of 5 | 19 Feb 2026 |
Not ranked yet
These models have independent or third-party results on too few of the 5 benchmarks to rank.
