GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation.
Benchmarks
| Benchmark | Category | Score | Variant | Provenance | Source | Observed |
|---|---|---|---|---|---|---|
| AA Agentic Index | agentic | 47.8 | non-reasoning | independent | aa | Sep 9, 2026 |
| 51.0 | — | mirror | openrouter | Sep 20, 2026 | ||
| 50.2 | xhigh-effort | independent | aa | Sep 20, 2026 | ||
| 51.0 | — | independent | aa | Sep 20, 2026 | ||
| AA Coding Index | coding | 76.9 | — | mirror | openrouter | Sep 5, 2026 |
| 75.9 | xhigh-effort | independent | aa | Sep 5, 2026 | ||
| 76.2 | non-reasoning | independent | aa | Sep 5, 2026 | ||
| 76.9 | — | independent | aa | Sep 5, 2026 | ||
| AA Intelligence Index | intelligence | 45.2 | non-reasoning | independent | aa | Sep 9, 2026 |
| 55.3 | xhigh-effort | independent | aa | Oct 28, 2026 | ||
| 56.9 | — | independent | aa | Oct 28, 2026 | ||
| 56.9 | — | mirror | openrouter | Oct 28, 2026 | ||
| Median output speed | speed | 47 tok/s | xhigh-effort | independent | aa | Nov 19, 2026 |
| 44 tok/s | — | independent | aa | Nov 19, 2026 | ||
| Median time to first token | speed | 268.36s | xhigh-effort | independent | aa | Nov 19, 2026 |
| 382.05s | — | independent | aa | Nov 19, 2026 |
Includes index data from Artificial Analysis.
Providers
Live operational stats per hosting endpoint, via OpenRouter.
| Provider | Quant | Context | $/1M in · out | Uptime 24h | Latency | Throughput |
|---|---|---|---|---|---|---|
| Amazon Bedrock | — | 1.1M | $10 · $50 | 99.32% | — | — |
| Azure | — | 1.1M | $10 · $50 | 100.00% | — | — |
| OpenAI | — | 1.1M | $10 · $50 | 99.93% | — | — |
| OpenAI | — | 1.1M | $20 · $100 | 100.00% | — | — |
| OpenAI | — | 1.1M | $5 · $25 | 100.00% | — | — |
| OpenAI | — | 1.1M | $60 · $300 | — | — | — |