GPT-6.2 Astra and GPT-6.2 Sol Arrive in Codex, With Ultrafast Mode for Pro Subscribers

CodexView original changelog

OpenAI made GPT-6.2 Astra and GPT-6.2 Sol available in ChatGPT Work and Codex. Astra is positioned for the hardest projects, such as changing and testing across an existing codebase, while Sol targets everyday coding, writing and multistep workflows and completes more tasks than GPT-6.1 Sol on the same paid plan. Ultrafast mode is available to Pro subscribers, with Sol on the $200 tier and both models on the $500 tier.

Key Takeaways

  • GPT-6.2 Astra and GPT-6.2 Sol are now both available in ChatGPT Work and Codex, giving developers two distinct model tiers.
  • Astra is aimed at the hardest projects, such as implementing and testing a change across an existing codebase.
  • Sol is the everyday model, recommended for coding, writing and multistep workflows.
  • On the same paid plan, Sol completes more tasks than GPT-6.1 Sol, so existing subscribers get more output without upgrading.
  • Ultrafast mode is available with Sol for Pro $200 subscribers.
  • Pro $500 subscribers can use Ultrafast with both Sol and Astra, making speed a plan-gated feature.

Two New Models in Codex

OpenAI added GPT-6.2 Astra and GPT-6.2 Sol to ChatGPT Work and Codex. Both models can be selected in Codex, and OpenAI points developers to its Codex model guidance for choosing between them.

Which Model to Pick

GPT-6.2 Astra for the Hardest Work

OpenAI recommends GPT-6.2 Astra for the most demanding projects. The examples it gives include implementing and testing a change across an existing codebase, and turning a report into slides that follow a given template. In Codex terms, this is the model to reach for when a task spans many files and needs careful verification.

GPT-6.2 Sol for Everyday Tasks

GPT-6.2 Sol is recommended for everyday coding, writing and multistep workflows. OpenAI states that on the same paid plan, Sol completes more tasks than GPT-6.1 Sol did, which means developers get more finished work out of the same subscription without changing plans.

Ultrafast Mode

Speed options are tied to the subscription tier. Pro subscribers on the $200 plan can run GPT-6.2 Sol in Ultrafast mode. Pro subscribers on the $500 plan can use Ultrafast with both GPT-6.2 Sol and GPT-6.2 Astra. Plan and workspace requirements for speed options are documented in the Codex speed configuration docs.

What This Means for Codex Users

The release gives Codex users a clear split between a top-end model for large, risky changes and a more efficient default for daily work. Developers who already use GPT-6.1 Sol can switch to GPT-6.2 Sol for a throughput gain, while those on higher Pro tiers can pair either model with Ultrafast mode.