Model · OpenAI · Text
GPT-6 Astra
OpenAI's most capable model, built for the hardest end-to-end work — computer use, browsing, software engineering, and scientific analysis. The GPT-6 generation's flagship, sitting above the GPT-5.6 tiers.
- Modality
- Text
- License
- Proprietary (Proprietary)
- Context window
- 1,050,000 tokens 1.05M-token context with 128k max output. Knowledge cutoff 2026-04-30. Reasoning effort runs low / medium / high / xhigh / max — the `none` level is not supported.
- Released
- September 3, 2026
- Last verified
- September 8, 2026
- Runs locally
- No
- Also handles
- vision, reasoning, code
Strengths
- State-of-the-art on computer use, browsing, software engineering, cybersecurity, and science
- 1.05M-token context with 128k max output and a 2026-04-30 knowledge cutoff
- Broad hosted tool support — web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP, and tool search
- Responses API adds async tool calling, mid-turn steering over WebSockets, and changing reasoning effort mid-conversation while keeping the cached prefix
Weaknesses
- Flagship pricing ($10 input / $50 output per MTok; $1 cached input)
- Tool calling requires the Responses API — Chat Completions cannot drive its tools
- No custom `temperature` / `top_p`, no log probabilities, and no `none` reasoning effort
- First model to reach the Critical cybersecurity level under OpenAI's Preparedness Framework, so the public model refuses advanced cyber tasks such as writing proof-of-concept exploits
- Not available for Realtime, Assistants, fine-tuning, embeddings, image generation, video, speech, transcription, translation, or moderation
- Closed weights — API and product only
Try it
| Where | Type | Notes |
|---|---|---|
| OpenAI API | hosted-api | API key required |
Used in solutions
Official sources
- Model docs docs
- API changelog (release note) announcement
Change log
- — Initial entry. Released 2026-09-03 per OpenAI's API changelog as 'our most capable model, built for the hardest end-to-end work'. Specs and pricing verified against the gpt-6-astra model docs.
Esc