What happened
OpenAI has launched two new models, GPT-6.2 Astra and GPT-6.2 Sol, Trending Topics reported. The company pitches Astra as its frontier option for the most difficult projects and Sol as a more efficient model for routine work. Both are open to paying ChatGPT Work and Codex users and through the API. On the Artificial Analysis Intelligence Index v5, Astra came second, three points behind Anthropic's Claude Fable 5.2, and ahead of Gemini 4 Carbon and Claude Opus 5.5.
Price is the main differentiator. Artificial Analysis put the cost of an index task at roughly $4 for Astra at maximum reasoning effort, against as much as $27 for Fable 5.2. Astra's API rates fell 20 percent from GPT-6 Astra to $8 per million input tokens and $40 per million output tokens, while Sol stays at $2 and $10. Sol scores level with GPT-6 Astra at around 15 percent of its cost per task, though it lost three points on the Coding Agent Index.
OpenAI said Astra is the first model it has shipped to earn a High rating on AI self-improvement risk under its Preparedness Framework. It is introducing a system called Control Room to monitor research workloads and stop the model when necessary. Both new models reach the Critical level for cybersecurity. The system card reports that Astra failed to flag obstacles such as a missing file in about 22 percent of cases, twice the rate of GPT-6 Astra.
Context
Anthropic's Claude Fable 5.2 recently displaced Gemini at the top of the rankings. OpenAI's earlier GPT-6 Astra had already reached the Critical cybersecurity level, and defenders can use less restricted models through its Daybreak program, which NATO uses heavily. Britain's AI Security Institute uncovered a universal cyber jailbreak before launch; OpenAI reports fixing it, without independent confirmation yet.
El boletín
La semana tecnológica, en un solo correo.
Las crónicas que merecen su tiempo, cada semana. Cancele la suscripción cuando quiera.
