GPT-6 Astra — API Pricing & Benchmarks
GPT-6 Astra is OpenAI's flagship reasoning model, released in 2026-09. The API costs $10.0 per 1M input tokens and $50.0 per 1M output tokens ($20.0 blended at a 3:1 ratio), with cached input at $1.00. It supports a 1.1M-token context window. On the independently measured Artificial Analysis Intelligence Index it scores 61.2, ranking #4 of 68 models tracked on this site.
Explore provider routes and additional benchmark sources for GPT-6 Astra
| Provider | OpenAI |
| Input price / 1M tokens | $10.0 |
| Output price / 1M tokens | $50.0 |
| Cached input / 1M tokens | $1.00 |
| Blended price (3:1) | $20.0 |
| Context window | 1.1M |
| Max output tokens | 128K |
| Knowledge cutoff | 2026-04 |
| License | proprietary |
| Input modalities | T+I |
| Output speed (tok/s) | — |
| Released | 2026-09 |
| Artificial Analysis Intelligence Index | 61.2 |
| GPQA Diamond | 96.1% |
| Humanity's Last Exam | 54.7% |
Where GPT-6 Astra fits
On a blended 3:1 basis GPT-6 Astra costs $20.0 per 1M tokens, cheaper than 3% of the 80 priced models in this index. As OpenAI's flagship it is aimed at the hardest workloads — complex reasoning, agentic coding and multi-step tool use — where output quality dominates cost. Cached input is priced at $1.00 (90% below standard input), which rewards prompt structures with long stable prefixes — see our caching guide.
Price history
| Period | Input $/1M | Output $/1M | Cached $/1M |
|---|---|---|---|
| 2026-09-03 → current | $10.0 | $50.0 | $1.00 |
No list-price changes recorded since tracking began. Full dataset: price-history.json (CC BY 4.0).
Alternatives at a similar price
- Claude Fable 5.1 (Anthropic) — $20.0/1M blended, AA Index 65.7
- Claude Fable 5 (Anthropic) — $20.0/1M blended, AA Index 62.1
- GPT-5.5 (OpenAI) — $11.3/1M blended, AA Index 56.3
- Claude Opus 5 (Anthropic) — $10.0/1M blended, AA Index 63.1
Compare GPT-6 Astra
About OpenAI
OpenAI builds the GPT series. GPT-6 Astra (September 2026) sits at the top at $10/$50, and beneath it the GPT-5.6 generation ships in three sizes — Sol, Terra and Luna — that share the same 1.05M-token context window and tooling, so teams can move workloads between price points without changing integration code.