GPT-6.2 Astra and Sol
OpenAI released GPT-6.2 Astra, its new frontier model, and GPT-6.2 Sol, an efficient model, with Astra API prices 20% below GPT-6 Astra.
Both models give better results on demanding tasks with fewer output tokens. Astra is trained to produce slides, spreadsheets and documents that follow professional conventions, and scores 9 points above GPT-6 Astra on Terminal-Bench-Science 0.1 at under one-third the cost. Sol keeps GPT-6.1 Sol's prices and, in Cognition's FrontierCode 2.0, comes within 0.5 points of Astra at about one-fifth the cost.
- Date
- Tuesday 17 November 2026
- Lab
- OpenAI
- Kind
- model
- Access
- closed API
- Price
- GPT-6.2 Astra $8 input, $0.40 cached input, $40 output per 1M tokens; GPT-6.2 Sol keeps GPT-6.1 Sol's prices ($2 input, $10 output per 1M per 9to5Mac)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Terminal-Bench-Science 0.1 gain over GPT-6 Astra | 9 percentage points at less than one-third the cost | company |
| FrontierCode 2.0, Sol vs Astra gap | within 0.5 points both at max, Sol at about one-fifth the cost | Cognition |
| Astra API input price | $8 per 1M tokens cached input $0.40 | company |
| Astra API output price | $40 per 1M tokens 20% lower than GPT-6 Astra | company |
Astra is OpenAI's first externally released model rated High for AI self-improvement, and both models meet Critical for cybersecurity and High for bio and chem under the Preparedness Framework. Also available in ChatGPT Work and Codex for paid plans.
Sources
- openai.com/index/introducing-gpt-6-2-astra-sol
- developers.openai.com/api/docs/changelog
- techmeme.com/261117/p49#a261117p49
- latent.space/p/ainews-gpt-6-2-astra-the-2-frontier-model-is-better-and-20-600-cheaper-but-
- technology.org/2026/11/18/openai-gpt-6-2-astra-sol-high-self-improvement-pricing/
This record was partly confirmed: some claims could not be checked on 17 November 2026. How we check