Migrating: GPT-5.5 to GPT-6 Sol

This is the move the GPT-5.5 retirement forces. On October 14, 2026, GPT-5.5 leaves ChatGPT, ChatGPT Work and Codex on every plan, and for Codex with ChatGPT sign-in on Plus, Pro, Business, Enterprise and Edu, GPT-6 Sol is the replacement OpenAI names. API users aren't forced to move, but the table above makes a case for it anyway.

Nothing you sized around changes

Window, maximum output, long-context rule and the full rate-limit table — batch-queue limits included — carry over unchanged. Throughput planning done for GPT-5.5 holds on GPT-6 Sol.

Every price drops

Input, cached input and output are all cheaper on GPT-6 Sol, with output falling furthest. The one structural change is caching: GPT-5.5 has no separate write rate, so written tokens bill at its input rate; GPT-6 Sol bills writes at its own cache-write rate. That rate is still lower than GPT-5.5's input rate, so the new line item costs less than what it replaces.

What to change

• Model id: gpt-5.5 becomes gpt-6-sol in config.toml, every profile, custom agent, scheduled task and script.
• Reasoning effort: your current setting still works — GPT-6 Sol accepts none through xhigh as before, plus max — but OpenAI says efforts don't map exactly between generations, so compare your setting with one level lower.
• API caching: replace prompt_cache_retention with prompt_cache_options.ttl, and review the new cache boundaries.
• Enterprise workspaces: an administrator must enable GPT-6 Sol before anyone can select it.

The alternative worth testing

GPT-6.1 Sol, the CLI's default, matches GPT-6 Sol on input and output price and is cheaper on cached input, but doesn't accept none. If your work is cache-heavy and never switches reasoning off, test it too. The full GPT-5.5 checklist covers every place the old id hides.