How to choose GPT-6.2 Astra or Sol in Codex
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol in ChatGPT Work and Codex on November 17. The models page recommends Astra for demanding work such as tracing a bug across services, fixing it and testing the change, and Sol for everyday work like editing code or drafting documents. At Standard speed, Sol costs fewer credits per token than Astra.
- Start the Codex CLI with the model you want.
codex -m gpt-6.2-astra
codex -m gpt-6.2-sol
In a session that's already running, switch with /model.
- The same option works for non-interactive runs.
codex exec -m gpt-6.2-sol "Review the current changes"
- To use it by default, set it in your config file.
# ~/.codex/config.toml
model = "gpt-6.2-sol"
The desktop app and the IDE extension read the same file. The models page lists the Codex app and CLI for GPT-6.2, so if you share the config with the IDE extension, it suggests picking a model every client has, such as gpt-6.1-sol.
The default context in Codex is 272,000 tokens for Astra and 500,000 for Sol. Both support up to 1 million tokens; set model_context_window in the config file to use more.
model_context_window = 1000000
In the desktop app and ChatGPT Work on the web, open Advanced to pick a model. New Work chats start with GPT-6.2 Sol at Light (Medium in Enterprise workspaces with the model enabled). Start from the default reasoning effort and adjust it to the task.
Ultrafast mode is available with GPT-6.2 Sol on the Pro $200 plan, and with both Sol and Astra on the Pro $500 plan.
Sources
Get notified when a new tip lands
Tip alerts only go out in the app. Install it and turn notifications on.