AI News
All stories
Official

OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol in its API

OpenAI's API changelog adds GPT-6.2 Astra for demanding reasoning and coding and GPT-6.2 Sol for everyday coding and agent work.

Models · · Updated · 2 min read

OpenAI added GPT-6.2 Astra and GPT-6.2 Sol to its API on November 17, according to the API changelog. The changelog positions Astra for demanding reasoning and coding tasks and Sol for everyday coding and agent workflows. Both are available through the Responses API and Chat Completions. The model pages draw the same split in fuller terms: Astra is described as the most capable model for the hardest projects, while Sol is described as the go-to model for everyday work, from coding to powering agents at scale.

Price is the clearest difference. Astra costs $8 per million input tokens and $40 per million output tokens, with cached input at $0.40 and cache writes at $10. Sol costs $2 input and $10 output, with cached input at $0.10 and cache writes at $2.50. Both models list a 1,050,000-token context window, a 922,000-token maximum input, and 128,000 maximum output tokens. Astra's knowledge cutoff is May 18, 2026, and Sol's is Jun 01, 2026. Both pages rate reasoning support as Highest and speed as Fast, so those labels do not tell the two apart.

Both models accept text and image input and produce text output. Neither supports audio or video. Both support streaming, function calling, and structured outputs, but not fine-tuning. The Responses API offers the wider toolset, including web search, file search, image generation, code interpreter, computer use, MCP, and tool search. According to the changelog, tool calling requires the Responses API, so an existing Chat Completions integration that relies on tools will have to move before it can use them. Both pages also mention the Multi-agent beta, which lets the Responses API delegate work to subagents.

Developers who turn reasoning off will run into a restriction on both models: the pages say the API rejects requests that set reasoning effort to none. Both pages also say the Responses API supports discounted background completion windows in public beta. Batch processing costs 50% of Standard rates. Fast mode costs twice the Standard rate for the same context length. Ultrafast costs 3 times Standard through February 28, 2027, and 6 times Standard starting March 1, 2027, Pacific time.

What changed

GPT-6.2 Astra and GPT-6.2 Sol were added to the OpenAI API, and tool calling now requires the Responses API rather than Chat Completions.

Type: New capability · Applies to: GPT-6.2 Astra, GPT-6.2 Sol, Responses API, Chat Completions

Why it matters

Teams that build products on OpenAI models now have two new options, and any integration that relies on tool calling has to run on the Responses API.

What you can do

Developers using OpenAI models can review the GPT-6.2 migration guide and confirm their integration uses the Responses API if it depends on tool calling.

How the story developed

  1. This coverage adds that the models are rolling out across ChatGPT and Codex as well as the API, with Astra positioned as the more capable option and Sol as the lower-cost default.

  2. Adds Artificial Analysis and Vals cost and benchmark figures, Sol's pricing and the mixed early tester reactions.

  3. Adds migration details: changed reasoning preservation when editing conversation prefixes, changed handling of the instructions field, and the requirement to use the Responses API for tool calling.

  4. This write-up adds the published per-token prices, the cached-input price, the Sol price comparison with earlier models, and early mixed reviews.

  5. Independent benchmarks add Artificial Analysis's Intelligence Index ranking, pricing comparisons, coding agent results and hallucination rates for both models.

  6. The launch post adds availability across ChatGPT Work, Codex and the API. It also claims gains in professional outputs, computer use and cost efficiency.

Original source

Further reporting

Where we saw it

6 sources

Artificial Analysis, Forward Future (Matthew Berman), Latent Space, OpenAI, OpenAI API changelog, Simon Willison

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works