·4 min read
GPT-6.2 Astra: Coding, Professional Work, and API Pricing

OpenAI introduced GPT-6.2 Astra on November 17, 2026, positioning it as a reasoning model for complex coding and professional work. Developers can use gpt-6.2-astra in the API. Teams considering an upgrade should compare quality, time, and full cost on their own repositories, documents, and workflows before rolling it out. OpenAI announcement
What is GPT-6.2 Astra designed for?
OpenAI positions Astra for demanding tasks such as tracing bugs across services, building and revising software, producing documents that follow templates, and complex engineering work. It accepts text and image input and directly outputs text. It can use a computer or work with files in supported tool environments, but image input does not mean the model directly outputs images. Model documentation
Start with a task that has clear acceptance criteria, such as reproducing a bug, proposing a fix, and running existing tests. Generated code, spreadsheets, and designs still need human review and verification. This article did not test the model.
How should you choose between Astra and Sol?
OpenAI also introduced GPT-6.2 Sol for everyday coding and agent workflows; Astra targets harder reasoning and coding. This is OpenAI’s positioning, not a guarantee that Astra is more economical for every task. Compare them using the same inputs and acceptance criteria, and measure first-pass success, human correction time, and cost per successful task. API changelog
API specifications and pricing
The model page lists a 1,050,000-token context window, a maximum of 922,000 input tokens, and 128,000 output tokens. The model directly supports text input and output and image input; it does not directly support audio or video. These are capacity and modality limits, not guarantees of accurate retrieval from long documents.
Standard short-context text pricing is $8 per million input tokens, $0.40 per million cached input tokens, and $40 per million output tokens; a 30-minute cache write costs $10 per million tokens. When input exceeds 272,000 tokens, input and cache rates double and output is charged at 1.5 times the short-context rate for the entire request. Batch and other processing tiers have different rates; tools can add costs. Check current model pricing and conditions before use.
For example, a standard short-context task with 100,000 regular input tokens and 10,000 output tokens costs $0.80 + $0.40 = $1.20 in model-token charges. This is an arithmetic example, not a measured bill, and excludes cache writes, tools, and other fees. OpenAI says standard input and output token prices are 20% lower than GPT-6 Astra; total task costs still depend on usage and retries. Launch announcement
What should teams check before migration?
- Use model ID gpt-6.2-astra and check account access and endpoint support. OpenAI lists Responses and Chat Completions; follow the Responses API documentation for tool calling.
- Review reasoning settings: the model documentation says requests setting reasoning effort to none are rejected. Check configuration and error handling in existing integrations.
- Prepare representative cases with expected answers or test results. Compare the current model and Astra using the same tools, data, and acceptance criteria; record quality, latency, and total token usage.
- Start in a test environment with restricted file, network, and write permissions. Keep appropriate human review for releases, data changes, and other consequential actions before expanding.
The model page lists Responses API tools including web search, file search, code execution, and computer use. Check documentation and account settings for actual availability and limits. Model tools
How should you interpret benchmarks?
OpenAI reports professional-work, coding, and scientific benchmarks and says the new model achieved better results at lower task cost on some tests. Results depend on datasets, model settings, and tool environments and may not transfer to every team. Evaluate reliability, review time, and full cost together on your own tasks.
Frequently asked questions
Where is GPT-6.2 Astra available?
OpenAI announced GPT-6.2 availability in ChatGPT Work, Codex, and the API. Check the actual product and official guidance for plan, account, and usage limits.
Does it directly generate images or video?
The model directly accepts image input and outputs text; it does not directly support video input or output. A supported image-generation tool can be used in the Responses API, depending on tools and configuration.
Is 272,000 tokens the context limit?
No. It is the published long-context pricing threshold. Context capacity and maximum input are listed separately.
Does a 20% price reduction mean each task costs 20% less?
Not necessarily. Input and output volumes, caching, long-context rates, tool calls, and retries all affect the final cost.
Official sources
- Introducing GPT-6.2 Astra and Sol, November 17, 2026.
- GPT-6.2 Astra model and pricing.
- OpenAI API changelog.
Verified November 18, 2026. Pricing, terms, and availability may change.