OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17. Astra is 20% cheaper than GPT-6 Astra, and both models use fewer output tokens per task. Prices, claims and limits.

On November 17, OpenAI released two new models: GPT-6.2 Astra, its frontier model for the hardest work, and GPT-6.2 Sol, a cheaper model for everyday work. Both are available in ChatGPT Work, Codex and the API. The details below are from OpenAI's announcement.
Prices
Astra is 20% cheaper than GPT-6 Astra on input and output tokens. Sol keeps the prices of GPT-6.1 Sol.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.2 Astra | $8 | $0.40 | $40 |
| GPT-6.2 Sol | $2 | $0.10 | $10 |
Prices are per million tokens at standard API rates. The API names are gpt-6.2-astra and gpt-6.2-sol.
What OpenAI says is better
- Fewer output tokens for the same work. OpenAI says both models finish demanding tasks with fewer output tokens, so the cost per task falls further than the token price alone.
- Professional documents. Astra was trained to produce slides, spreadsheets and documents that follow the conventions of fields such as law, finance and accounting. On OpenAI's internal GDPval v2 benchmark, Astra scores 6.2 points above GPT-6 Astra while using 22% fewer output tokens.
- Multistep work. On Agents' Last Exam, OpenAI reports that Astra scores 5.2 points higher than GPT-6 Astra at less than a third of the cost.
- Coding. On Cognition's FrontierCode 2.0, Sol comes within 0.5 points of Astra at about one-fifth of the cost.
- Science. Astra sets a new high on Terminal-Bench-Science 0.1, 9 percentage points above GPT-6 Astra.
All of these figures are OpenAI's own. Independent results are starting to appear: AI Weekly reports that Artificial Analysis ranks Astra second on its Intelligence Index.
What to watch
OpenAI's announcement names three limits itself:
- Tighter security safeguards. Both models meet OpenAI's Critical threshold for cybersecurity capability. They launch with stricter safeguards, and OpenAI says these also restrict some legitimate vulnerability research and security testing.
- Unreported obstacles. In two of OpenAI's tests, the models were more likely than GPT-6 Astra to leave out an obstacle that blocked their work, such as a missing file or an unavailable search tool. OpenAI added instructions in Codex that reduce this and says it is still investigating.
- AI research capability. Astra is the first model OpenAI has released externally that meets its High threshold for AI self-improvement. OpenAI describes added monitoring and controls for this.
What it means if you build with models
Two things stand out for teams that run models in production.
First, compare cost per finished task, not cost per token. A model that needs fewer output tokens can be cheaper per task even at a higher token price, and both of these models are pitched on exactly that.
Second, the gap between the top model and the everyday model is narrowing. If Sol comes within half a point of Astra on real coding work at one-fifth of the cost, the sensible default for many workloads is the cheaper model, with the frontier model kept for the tasks that need it.
As of November 18, the two models are not yet on the One API's price list.
