Skip to main content

GPT-6.2 Astra

Overview​

GPT-6.2 Astra is OpenAI's GPT-6.2 reasoning model for demanding software engineering and professional projects, released on November 17, 2026. It combines text and image understanding, adjustable reasoning, and tool use to plan and complete complex work.

Key Features​

  • Complex Project Work: Plans and carries out projects, adapting its approach as new information becomes available.
  • Software Engineering: Builds and revises software, using code execution and editing tools to support implementation and verification.
  • Visual Computer Use: Interprets screenshots and works with computer-use tools to carry out tasks across applications.
  • Long Working Context: Accommodates extensive source material, code, and conversation history within a 1.05M-token context window.
  • Agent Delegation: Supports the Responses API Multi-agent beta, allowing independent subtasks to be delegated to subagents.
  • Clear Writing: Produces concise, direct writing by default, with tone and detail adjustable for the intended audience.

Best Use Cases​

  • Complex Software Projects: Implementation, debugging, and code review that involve multiple components and repeated testing.
  • Research and Technical Decisions: Open-ended analysis that requires comparing evidence, testing assumptions, and weighing trade-offs.
  • Professional Deliverables: Turning a detailed brief and supporting material into reports, documents, or applications with iterative revision.
  • Computer-Based Workflows: Coordinating a sequence of actions across interfaces, files, and connected tools.

Capabilities and Limitations​

CapabilityDescription
ReasoningSupports low, medium, high, xhigh, and max; none and minimal are unsupported.
MultimodalText and image input; text output. Native audio and video are unsupported.
Context Window1,050,000 tokens, shared by input and generated tokens.
Max Input922,000 tokens.
Max Output128,000 tokens, including reasoning and visible output.
Tool UseSupports functions, web and file search, code execution, computer use, MCP, and image generation through Responses API tools.
Structured OutputSupports Structured Outputs and streaming.
Prompt CachingSupports implicit and explicit caching, with 30-minute or 12-hour minimum lifetimes in the Responses API.
Knowledge CutoffMay 18, 2026.

Known Limitations​

  • Tool calling requires the Responses API; Chat Completions supports requests without tools.
  • Sampling and log-probability controls are unsupported. Fine-tuning is unavailable.
  • Reasoning consumes the output budget and can increase latency; an insufficient budget can leave a response incomplete before a visible answer is produced.
  • Editing earlier conversation content or switching to a smaller model can discard prior reasoning, reducing continuity and increasing repeated work.

Standard Pricing​

The token prices below are shown in USD per 1 million tokens; web search is billed in USD per use.

ContextInput
(USD / 1M Tokens)
30m Cache Write
(USD / 1M Tokens)
12h Cache Write
(USD / 1M Tokens)
Cache Read
(USD / 1M Tokens)
Output
(USD / 1M Tokens)
Web Search
(USD / use)
Up to 272K input tokens$8.00$10.00$20.00$0.40$40.00$0.01
More than 272K input tokens$16.00$20.00$40.00$0.80$60.00$0.01
  • Above 272K input tokens, the higher rates apply to the entire request, not only the portion above the threshold.
  • Cache-write charges apply to both implicit and explicit caching and replace the ordinary input charge for those tokens. The default minimum lifetime is 30 minutes. A 30-minute write costs 1.25x the applicable input rate; a 12-hour write costs 2.5x. Both lifetimes have the same cache-read rate, 0.05x the applicable input rate.
  • Reasoning tokens are billed as output tokens.
  • Web search costs 10,000 Credits ($0.01) per use and is not affected by the context tier. Search content tokens are billed at the applicable model token rates.

Credits settlement: B.AI converts charges at 1 USD = 1,000,000 Credits and deducts Credits from the account balance.

Pricing note

Model prices in the documentation are based on the model providers' official list prices and serve as B.AI standard reference prices. B.AI may offer discounts or limited-time promotions on this basis; offer details are shown in Promotions and Pricing Updates and the website's discount groups. Applicable prices, offer terms, and final charges are subject to the platform display and actual billing records.