Wednesday, November 18, 2026 Editions
Latest
Trending
Tech

OpenAI Launches GPT-6.2 Astra and Sol in ChatGPT Work and Codex

OpenAI released GPT-6.2 Astra for demanding professional work and GPT-6.2 Sol for everyday work on Tuesday, with lower Astra pricing, higher benchmark scores, and the company's first High rating for AI self-improvement.

Read inENESNEHIFR
OpenAI Launches GPT-6.2 Astra and Sol in ChatGPT Work and Codex
Photo: HaeB, CC BY-SA 4.0, via Wikimedia Commons (https://commons.wikimedia.org/wiki/File:Pioneer_Building,_San_Francisco_(2019)_-3.jpg)

Key points

  • OpenAI released two new models on Tuesday, GPT-6.2 Astra and GPT-6.2 Sol.
  • Both models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users.
  • OpenAI split the release by workload.

OpenAI released two new models on Tuesday, GPT-6.2 Astra and GPT-6.2 Sol. Astra is the frontier model for demanding professional work. Sol is the efficient model for everyday work.

Both models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. Developers can use them through the API as gpt-6.2-astra and gpt-6.2-sol. The rollout does not add either model to regular ChatGPT chat.

Two models, two jobs

OpenAI split the release by workload. Astra targets hard professional tasks where accuracy matters most. Sol targets routine work where speed and cost matter more. Both models list a context window of 1,050,000 tokens, with a maximum of 922,000 input tokens. The knowledge cutoff is May 18, 2026 for Astra and June 1, 2026 for Sol.

Pricing and performance

Astra costs $8 per million input tokens and $40 per million output tokens through the API. That output and input pricing is a 20 percent cut from GPT-6 Astra. Cached input costs $0.40 per million tokens. Sol keeps the GPT-6.1 Sol price of $2 per million input tokens and $10 per million output tokens.

OpenAI says Astra scores 5.2 points higher than GPT-6 Astra on its Agents' Last Exam benchmark. The company also says Astra costs less than one third as much per task as the older model on that test. On OpenAI's internal ExploitBench cybersecurity benchmark, Astra jumps 22 points over GPT-6 Astra. OpenAI calls Astra its most capable model ever for cyber work.

Independent evaluator Artificial Analysis ranks GPT-6.2 Astra second on its Intelligence Index. Anthropic's Claude Fable 5.2 ranks first.

Safety ratings and a cancelled predecessor

The GPT-6.2 system card rates both models Critical for cybersecurity under OpenAI's Preparedness Framework. It rates both models High for biological and chemical capabilities. Astra is OpenAI's first externally released model rated High for AI self-improvement.

The launch follows a cancelled release. OpenAI cancelled GPT-6.1 Astra about two months earlier after internal tests raised safety concerns. The company says the new models launch with tighter safeguards for higher-risk cybersecurity assistance.

What changes in December

From December, Plus, Pro and Business plans will share one token-measured allowance across ChatGPT and Codex. Users on those plans will draw from the same allowance whether they work in ChatGPT or in Codex.

For professional users, the release pairs a stronger frontier model with a lower price per task. For OpenAI, the safety ratings set the terms of the launch: the same system card that records the benchmark gains also records Critical and High capability ratings, and the safeguards that ship with the models.

Timeline

  1. Nov 17, 2026OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol in ChatGPT Work and Codex, and through the API.
  2. December 2026Plus, Pro and Business plans will share one token-measured allowance across ChatGPT and Codex.

Sources