GPT-5.3-Codex vs Claude Sonnet 4.6
Côte à côte. Chaque métrique. Chaque benchmark.
| Type | GPT-5.3-Codex | Claude Sonnet 4.6 |
|---|---|---|
| Provider |  OpenAI |  Anthropic |
| score moyen | 52.2 | 47.6 |
| Prix d'entrée | $1.75 | $3.00 |
| Prix de sortie | $14.00 | $15.00 |
| Fenêtre de contexte | 400K tokens (~200 books) | 1.0M tokens (~500 books) |
| Sorti le | 2026-02-24 | 2026-02-17 |
| Code source ouvert | Proprietary | Proprietary |
Scores de benchmark
7 benchmarks · GPT-5.3-Codex: 5, Claude Sonnet 4.6: 2
| Benchmark | Catégorie | GPT-5.3-Codex | Claude Sonnet 4.6 |
|---|---|---|---|
| Artificial Analysis — Agentic Index | speed | 62.2 | 63.0 |
| Artificial Analysis — Coding Index | speed | 53.1 | 50.9 |
| Artificial Analysis — Quality Index | speed | 54.0 | 51.7 |
| PostTrainBench | knowledge | 17.8 | 16.4 |
| SWE Atlas — Codebase QnA | agentic | 32.6 | 31.2 |
| SWE-Bench verified | coding | 74.8 | 75.2 |
| WeirdML | coding | 79.3 | 66.1 |