OpenAI releases GPT-6.2 Astra for complex work and Sol as lower-cost model
Android Authority reported that OpenAI has launched GPT-6.2 Astra, which the company aims at complex professional work, scientific research and engineering, alongside GPT-6.2 Sol, its lower-cost option for everyday writing and multi-step assignments. Both models became available on November 17, 2026 to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and developers can call them through the API as gpt-6.2-astra and gpt-6.2-sol. Early testers said Astra flagged problems other models missed and proposed unconventional solutions. OpenAI has acknowledged that some reviewers missed late improvements and that the user experience still has room to improve.
Pricing is the sharpest change. Astra's API rates are 20 percent below its predecessor, at $8 per million input tokens, $0.40 for cached input and $40 for output, according to SiliconANGLE. Sol keeps the $2 input and $10 output rates of GPT-6.1 Sol. On OpenAI's GDPval v2 test, Astra gained 6.2 points over its predecessor while using 22 percent fewer output tokens. OpenAI also says Sol came within 0.5 points of Astra on Cognition's FrontierCode 2.0 coding benchmark at about one fifth of the cost. The system card places both models at OpenAI's Critical cybersecurity capability threshold, which brings tighter limits on higher-risk security assistance.
Why it matters
For teams building products, the pair turns model choice into a routing problem. Sol's token prices are roughly a quarter of Astra's, and OpenAI says it nearly matches Astra on coding, which makes high-volume agent loops and drafting natural candidates for the cheaper default. Astra fits work where errors are costly, such as legal drafting, scientific analysis and client-ready financial output. Teams should compare cost per completed task rather than cost per token before changing defaults.
For investors and sellers, the release shows price pressure at the top tier, with a 20 percent cut for Astra inside a single generation. The cyber classification and the state lawsuit add compliance work for any product that gives agents access to tools, code or external systems. Buyers in regulated sectors will ask how vendors handle incidents and test agent behavior, so companies that can document those controls will have an edge in procurement.
Key points
- 1Astra API prices are $8 per million input tokens and $40 per million output tokens, 20 percent below GPT-6 Astra.
- 2Sol keeps $2 input and $10 output pricing per million tokens, unchanged from GPT-6.1 Sol.
- 3OpenAI says Sol lands within 0.5 points of Astra on FrontierCode 2.0 at about one fifth of the cost.
- 4Vals AI measured Sol at $3.18 per task versus $10.68 for Astra, with both models ranking in its top 10.
- 5