Wednesday, 18 November 2026, story 1 of 7
OpenAI releases GPT-6.2 Astra at 20% lower API price, Sol unchanged
Both models are in the API; independent tests rank Astra second on one index but show flat coding scores and weak legal results.

What happened
OpenAI has released GPT-6.2 Astra, a frontier model, and GPT-6.2 Sol, an efficient model. Both are in the API as gpt-6.2-astra and gpt-6.2-sol, according to Simon Willison. The Decoder reports they are also available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.
Astra costs $8 per million input tokens, $0.40 for cached input and $40 per million output tokens. OpenAI says that is 20 percent below its predecessor. Sol keeps GPT-6.1 Sol's $2, $0.10 and $10. On Agents' Last Exam, OpenAI says Astra scores 5.2 points above GPT-6 Astra at less than a third of the cost.
What it means
The rate card is only part of the picture. Willison notes Astra's list price matches Claude Fable 5.2, so OpenAI's cost-per-task claims likely rest on token use as well as price. Vals AI reports Astra uses the fewest output tokens in the top 10. Builders should measure cost per task on their own workloads, not compare per-token prices.
The capability gains are uneven. Artificial Analysis finds coding flat, with Sol down three points on its Coding Agent Index. Vals AI ranks both models well down its legal benchmarks. Epoch AI found Astra weaker than Fable 5.2 on real-world research tasks. Teams using Sol for coding or Astra for legal work have reason to re-test before switching. Artificial Analysis also reports high hallucination rates, though on unusually hard questions. The $0.40 cached rate may matter most for long agent sessions.
What to do: Re-run your own coding and legal tasks on gpt-6.2-astra and gpt-6.2-sol, and compare cost per task, before moving production traffic from the previous models.
Sources
- OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains, The Decoder
- GPT‑6.2 Astra and GPT‑6.2 Sol, Simon Willison
Filed under Models. Published .
"What it means" is analysis, not reporting. Found an error? Write to daily@vibitely.com.