GitSeonjeap
English

← News

Industry

OpenAI Releases GPT-6.2 Astra and Sol, Publishes API Pricing

Posted As of

OpenAI has released two models: Astra, a frontier-class model, and Sol, a model for everyday work. They are available in ChatGPT Work, Codex, and the API, and per-million-token pricing is listed in a community announcement.

What happened

  • According to OpenAI's official blog, GPT-6.2 Astra is a frontier model for the most difficult projects, and GPT-6.2 Sol is an efficiency-focused model for everyday work.
  • Performance gains (official): Astra is said to improve on GPT-6 Astra by 6.2%p while using 22% fewer output tokens. Sol improved by 5.3%p over GPT-6.1 Sol.
  • API pricing (official, per 1 million tokens, from the OpenAI developer community announcement): Astra input $8.00 / output $40.00, Sol input $2.00 / output $10.00. Cached input is $0.40 for Astra and $0.10 for Sol.
  • Availability: Provided in ChatGPT Work and Codex to Plus, Pro, Enterprise, and Edu users.
  • API identifiers are gpt-6.2-astra and gpt-6.2-sol, and both models offer an Ultrafast option from launch.
  • Safety measures: OpenAI said it launched with stricter safeguards in high-risk cybersecurity support areas.
  • Not confirmed in official sources: whether free users have access, and the Astra price-cut claim and self-improvement benchmark claims that appear only in secondary reports, are not covered in this article.

Our take

These models are offered through the API and cloud services, and the official materials make no mention of local execution or weight release. For users running local AI on Mac, PC, or glasses, this does not directly apply; it can be seen as a way to compare per-token prices for those calling the API. How large the performance gains are in real work will only be known once independent verification results come out.

Sources and related posts

More in Industry

  1. Google Starts Gemini 4 Neon Rollout in Antigravity (antigravity.google)

    The Gemini 4 Neon rollout has started in Google Antigravity. Paid Google AI Pro and Ultra subscribers can use it in the desktop app, CLI, and IDE extensions, while pricing and context length are not in the official blog post.

    Industry| |1 sources |Our take included
  2. NVIDIA Announces Rubin HPC GPU for Scientific Computing (blogs.nvidia.com)

    NVIDIA announced Rubin HPC, a GPU for HPC within the Vera Rubin platform, at SC26. Its expected launch is in 2028, and it is not a product currently on sale.

    Industry| |1 sources |Our take included
  3. NVIDIA Board Creates National Security and Export Control Oversight Committee (globenewswire.com)

    NVIDIA's board has created a standing committee to oversee national security policy, export control compliance, and global trade risk management. Director Suzanne Nora Johnson has been appointed chair of the committee.

    Industry| |1 sources |Our take included

All news