OpenAI has published a fuller account of GPT-6.2 Astra and Sol, describing gains in professional work, coding and scientific tasks. The most practical change in the detail: Astra’s API token prices are lower than GPT-6 Astra’s, while OpenAI presents Sol as a more capable everyday model at the same API prices as its predecessor.
OpenAI Watch analysis
What happened
OpenAI says Astra produces more polished professional documents and slides, writes cleaner code and handles complex changes to existing codebases. It says Sol is more efficient than GPT-6.1 Sol and can complete more tasks on the same paid plan. Both models, the company says, deliver better results on demanding tasks with fewer output tokens.
The company reports that Astra scored 5.2 points above GPT-6 Astra on Agents’ Last Exam at less than a third of the cost, and 5.5 points above Claude Fable 5.2 at 5% of the cost. OpenAI notes that the Fable result excludes nine tasks that used a fallback model. These are company-reported benchmark results, not independent comparisons.
For API users, OpenAI lists Astra input and output token prices as 20% lower than GPT-6 Astra’s, with a 60% discount on cached input: $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens. Sol keeps GPT-6.1 Sol’s API prices. OpenAI also reports improvements on benchmarks for science, legal work, coding and interactive 3D experiences.
Why it matters
The useful pitch is not simply “smarter models”. OpenAI is pairing claimed quality gains with lower Astra API costs and fewer output tokens, which could matter to developers running work at scale. For individual users, the company’s claims about better documents, spreadsheets, coding and multistep tasks suggest fewer rounds of polishing, though real-world results will depend on the job and the workflow.
The pricing distinction is worth keeping straight: Astra’s API prices fall, but Sol’s do not. And leaderboard scores, however tidy, are not a substitute for seeing how a model handles your actual work.
Our read
There is enough here to make the launch more than a model-name reshuffle. The combination of workflow claims, benchmark detail and a concrete Astra price cut gives users something testable. OpenAI’s results deserve interest, not a gold medal awarded by the company to itself: compare outputs and costs on your own tasks before changing a production workflow.
What to watch
- Whether independent evaluations reproduce the benchmark gains, including the reported cost differences.
- How Astra’s lower API prices translate into total costs for real workloads.
- Whether Sol’s claimed efficiency gains show up in everyday use at unchanged API prices.
Discussion spark: Would lower API costs and fewer output tokens persuade you to switch models, or do you need independent evidence of better results on your own work first?
Sources and evidence
- Source update (21 November 2026, 00:00 UTC)
OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.