Loading prices...
All news
Glossy violet crystal star on a stack of four gold coins next to a golden sun above a single gold coin, near-black violet background with a purple glow

OpenAI releases GPT-6.2 Astra and Sol, with Astra costing four times more

10:00 · 18.11.2026
Source: OpenAI
4

OpenAI released two GPT-6.2 models on Tuesday: Astra for the hardest professional and research work and Sol for everyday tasks. The API price for output differs fourfold: $40 per million tokens for Astra and $10 for Sol, according to OpenAI's announcement for developers.

Astra is 20% cheaper than GPT-6 Astra. Sol keeps the price of GPT-6.1 Sol.

Prices and access

  • GPT-6.2 Astra: $8 per million input tokens, $0.40 for cached input, $40 for output
  • GPT-6.2 Sol: $2 for input, $0.10 for cached input, $10 for output

Both models are in the API and are rolling out in ChatGPT Work and Codex on the Plus, Pro, Enterprise and Edu plans. A forum moderator confirmed that the rollout does not add them to regular ChatGPT chat.

Ultrafast mode generates tokens up to 4.5 times faster and uses up the plan's limit faster. Pro 500 subscribers get it with both models, and Pro 200 subscribers get it for the first time, starting with Sol. The API offers Ultrafast too.

“A frontier model built for your hardest projects.”

— OpenAI, on GPT-6.2 Astra

Quote source: OpenAI Developer Community, 17 November 2026.

What OpenAI claims and what testers found

OpenAI says Astra costs less than a third per task of GPT-6 Astra and beats Anthropic's Claude Fable 5.2 on Agents' Last Exam by 5.5 points. On FrontierCode 2.0, Sol trails Astra by half a point at about a fifth of the cost.

Independent results are mixed, The Decoder reported. On the Artificial Analysis Intelligence Index, Astra scores 62 and ranks second behind Claude Fable 5.2 at 65, while Sol scores 57. On very hard factual questions Astra produced an AI hallucination in 59% of answers and Sol in 55%. Vals AI ranks Astra 7th and Sol 9th, and both ranked low on legal tasks.

Epoch AI put Astra level with Claude Fable 5.2 on its capabilities index. On real research tasks, though, Astra trails Fable and barely improves on GPT-6 Astra.

Cyber limits

Both models reach the Critical cybersecurity level in OpenAI's Preparedness Framework, as GPT-6 Astra did. OpenAI launches them with tighter safeguards and says these will also block some legitimate security research. It is working to bring both models to Daybreak, its program for vetted defenders.

OpenAI also says Astra is the first released model to reach its High threshold for AI self-improvement, meaning it can speed up AI research, and lists that as a risk.

If you pay for API output, compare the two on your own tasks before switching: for routine work Sol costs a quarter as much and stays close on code.

Most benchmark figures come from OpenAI. This report is for information only.

Published: 10:00 · 18.11.2026
Maks

Author

Maks

Trading man

I've been interested in the cryptocurrency market for a long time, am a trader, and write articles and news about my experience and crypto in simple terms.

Comments (0)

No comments yet — be the first!

Join our community

No spam, unsubscribe in one click