Changelog
GPT-6.2 Astra and Sol replace GPT-6 Astra and GPT-6.1 Sol
OpenAI's GPT-6.2 Astra and GPT-6.2 Sol are in General Input, and they replace GPT-6 Astra and GPT-6.1 Sol everywhere. A specialist, workflow, app, or default model that names an older Astra or Sol now runs on its GPT-6.2 successor, so nothing has to be saved again. Auto's Balanced and Max tiers fall back to GPT-6.2 Sol, and members whose ChatGPT plan includes the new models run them on their own subscription through Token Tunnel. Compare them with every other model on the models page.
How it works
- GPT-6.2 Astra. $8 input and $40 output per million tokens, down from $10 and $50, and $0.40 per million cached input tokens, down from $1. It keeps Astra's 1.05M-token window and its low to max effort levels, defaulting to high.
- GPT-6.2 Sol. The same $2 input and $10 output per million tokens as GPT-6.1 Sol, with cached input halved to $0.10. It defaults to X-High. OpenAI no longer accepts the None level, so a saved pick that asked for None runs at Low, as OpenAI recommends.
- Long prompts. Past 272,000 input tokens, both bill input at twice the rate and output at 1.5 times, as before.
- Routes. OpenAI serves both first, with Vercel AI Gateway and OpenRouter behind it.
- Token Tunnel. Paid ChatGPT plans serve both. The tunnel plans for 500,000 tokens of context on GPT-6.2 Sol, up from 272,000 on 6.1 Sol, and 272,000 on GPT-6.2 Astra. On a subscription, Auto now picks GPT-6.2 Sol for Fast and GPT-6.2 Astra for Balanced and Max.
Also shipped
- On ChatGPT's free plan, which has no Sol, the model in Sol's place is now labeled GPT-5.6 Terra, the model it runs.