OpenAI releases GPT-6.2 Astra and Sol, cutting flagship price 20% to match Claude

OpenAI released GPT-6.2 Astra and Sol on November 17, cutting the flagship Astra's API price by 20% to match Anthropic's Claude Fable 5.2, while its safety card rates both models at the highest level for cybersecurity risk.
OpenAI released its new models GPT-6.2 Astra and GPT-6.2 Sol on Tuesday, November 17. Both are available now in ChatGPT Work, the coding tool Codex and the API for Plus, Pro, Enterprise and Edu users. The flagship Astra's API price has been cut by 20%, bringing it exactly in line with rival Anthropic's Claude Fable 5.2.
Two models, two jobs
Astra is aimed at the hardest professional work — OpenAI points to science, law and finance — while Sol is the lighter model for everyday tasks. Both models have a context window of 1.05 million tokens and a maximum output of 128,000 tokens.
Flagship 20% cheaper, Sol unchanged
Astra's standard API price is $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for GPT-6 Astra. Cached input falls to $0.40 per million tokens, 60% cheaper than before. Sol keeps its price of $2 in and $10 out per million tokens — a quarter of Astra's rate. OpenAI is also extending its faster Ultrafast mode to Pro 200 subscribers for the first time, starting with Sol.
Performance claims are OpenAI's own; independent tests differ
OpenAI says Astra beats its predecessor on several of its own benchmarks, but those results come from the company's own evaluations and early testers are split. The independent Artificial Analysis index scores Astra at 62, behind Claude Fable 5.2 at 65. Notably, Claude Fable 5.2 launched eight days before Astra at the identical list price of $8 in and $40 out per million tokens.
Safety card: top level for cybersecurity risk
OpenAI's system card, published the same day, rates both models at 'Critical' — the highest level — for cybersecurity risk under its Preparedness Framework, and 'High' for biological and chemical risks. Astra is also the first publicly released model to reach the 'High' threshold for AI self-improvement. OpenAI says the models passed a battery of safety tests before release, but whether outsiders can independently verify their safety remains to be seen.
(OpenAI's headquarters. File photo.)
N環球 
💬 Comments
No comments yet. Be the first!