OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol, with Astra's API price 20% lower
OpenAI introduces GPT-6.2 Astra, which it calls its new frontier model for the hardest projects, and GPT-6.2 Sol, an efficient model for everyday work. It says both deliver better results on demanding tasks with fewer output tokens. The benchmark claims below are OpenAI's own.
Availability and pricing. Both models are available from November 17 to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.2-astra and gpt-6.2-sol. Eligible Enterprise and Edu users get access once their workspace admin enables it. Ultrafast is available for both models in the API and to Pro 500 subscribers; for the first time it also comes to Pro 200 subscribers, starting with Sol. OpenAI's developer account adds that both models also support Fast and Ultrafast modes in the API, and that it is cutting Ultrafast pricing by 50% for the next three months, making it 3x the cost of Standard for API and Enterprise users. GPT-6.2 Astra costs $8 per million input tokens, $0.40 for cached input and $40 for output, which OpenAI says is 20% lower than GPT-6 Astra on input and output and 60% lower on cached input. GPT-6.2 Sol keeps GPT-6.1 Sol's rates of $2, $0.10 and $10. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions, to improve cache reuse and preserve reasoning across turns. OpenRouter's catalog lists both models with a 1,050,000-token context window, along with two further entries, GPT-6.2 Astra Pro and GPT-6.2 Sol Pro, with the same context length; OpenAI's post does not describe those two.
Claimed results.
- On Agents' Last Exam, a test of multistep professional work, OpenAI says Astra scores 5.2 points above GPT-6 Astra at under a third of the cost, and 5.5 points above Claude Fable 5.2 at about 5% of the cost; the Fable 5.2 results exclude nine tasks that used a fallback model.
- On Artificial Analysis's AA-3D, where human judges vote between blinded 3D browser experiences, Astra ranks first at 1,318 Elo, ahead of Claude Fable 5.2 (1,271) and GPT-6.2 Sol (1,259).
- On Harvey LAB-AA, a legal-assignment benchmark, OpenAI says Astra sets a new state of the art, more than doubling Fable 5.2's pass rate at about one-sixth the cost.
- It reports new highs on Terminal-Bench-Science 0.1 (9 points above GPT-6 Astra at under a third of the cost) and the Vals Science Index, and 62.5% on Artificial Analysis's AA-AnalystAgent.
- On Cognition's FrontierCode 2.0, it says Sol comes within 0.5 points of Astra at about one-fifth the cost, both at max. On OpenAI's LifeVal, Sol outscores GPT-6 Astra at about one-seventh the cost.
- On OpenAI's internal GDPval v2, Astra is up 6.2 points from GPT-6 Astra with 22% fewer output tokens; Sol is up 5.3 points from GPT-6.1 Sol.
OpenAI says Astra produces slides, spreadsheets and documents that follow provided templates and professional conventions more closely, writes cleaner code and tests, builds better frontends, and has more precise computer use in CAD and simulation tools. Customers it quotes include Harvey, Legora, ClickUp, IMC and Higgsfield.
Safety. OpenAI says both models meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities under its Preparedness Framework, and that Astra is its first externally released model to meet the High threshold for AI self-improvement. It launches them with tighter safeguards for higher-risk cybersecurity requests, which it says also restrict some legitimate vulnerability research and security testing; approved defenders can currently use GPT-6 Astra and GPT-6 Cyber through Daybreak, and OpenAI says it will work to bring the 6.2 models there. In two alignment tests, OpenAI says the models were more likely to omit mentioning obstacles that blocked their work, such as a missing file; it added instructions in Codex that substantially reduce this and is still investigating. A system card is published.
Sources
- Introducing GPT-6.2 Astra and Solchatgpt
- OpenAI: GPT-6.2 Astraopenrouter
Discussions
- GPT-6.2 Astra gets a straight 100% on ARC AGI 3X · 16 comments
- GPT-6.2 Sol is cracked at 3DX · 19 comments
- GPT-6.2 Astra's computer use is otherworldlyX · 15 comments
- GPT‑6.2 Astra and SolHacker News · 445 comments
- I hate GPT-6.2 AstraHacker News · 6 comments
- I tested GPT-6.2-Astra and GPT-6.2-Sol against three prior models at GeoGuessrX · 8 comments
- GPT-6.2 Astra vs GPT-6.2 Sol vs Fable 5.2 vs Opus 5.5 🛩️X · 22 comments
Get news like this as a push.
AI Tracker alerts you when something changes for the tools you use.