OpenAI retires GPT-5.5 from ChatGPT and Codex on 14 October 2026. Paid plans move to GPT-6 Sol, Free and Go to GPT-6 Luna, and the API keeps GPT-5.5.
OpenAI will retire GPT-5.5 from ChatGPT, ChatGPT Work and Codex on all plans on 14 October 2026. The OpenAI API is not affected, and GPT-5.5 stays available there. Anyone whose Codex sessions, scheduled tasks or scripts select gpt-5.5 through a ChatGPT sign-in needs to pick a new model.
Which model replaces it
OpenAI's models page names two replacements, both with the qualifier "when available":
| Plan | Replacement |
|---|---|
| Plus, Pro, Business, Enterprise, Edu | GPT-6 Sol (gpt-6-sol) |
| Free, Go | GPT-6 Luna (gpt-6-luna), in the desktop app |
The same page also recommends GPT-6.1 Sol as the primary model while the retirement notice names GPT-6 Sol. OpenAI does not explain the difference. Treat gpt-6-sol as the documented replacement.
OpenAI tells users to replace gpt-5.5 in workspace defaults, saved model settings, managed configurations, custom agents, scheduled tasks and scripts before the cutoff. The page's note on the earlier GPT-5.4 retirement says that Enterprise and Edu administrators must enable Luna first; the GPT-5.5 notice does not say whether that applies here.
What it means for Codex usage limits
The pricing page lists estimated local Codex messages per five-hour window for Plus and Standard Business. OpenAI stresses these are not fixed limits.
| Model | Est. local messages per 5 hours | Credits per 1M tokens (input / cached / output) |
|---|---|---|
| GPT-5.5 | not listed | 125 / 12.5 / 750 |
| GPT-6 Sol | 15 to 150 | 50 / 5 / 250 |
| GPT-6 Luna | 350 to 3,000 | 2.5 / 0.25 / 12.5 |
On credit rates, GPT-6 Sol is priced at 40 percent of GPT-5.5 on input (50 against 125) and a third on output (250 against 750). Teams that extend usage with ChatGPT credits should see lower burn per token on the same work, though real usage depends on how many tokens an agent run consumes. Luna is cheaper again, at 2.5 credits input and 12.5 output.
Codex credit rates per 1M output tokens. GPT-6 Sol costs a third of GPT-5.5; Luna far less. Source: OpenAI pricing docs.
Per plan:
- Free and Go: GPT-6 Luna at Standard speed in the desktop app, subject to rollout. No numeric limits are published.
- Plus: GPT-6.1 Sol and GPT-6 Luna, with credits available to extend usage.
- Pro: tiers at $100, $200 and $500 a month. OpenAI says Pro plans currently have no five-hour limit.
- Business: Standard seats share Plus's estimates.
- Enterprise and Edu: flexible pricing has no fixed rate limits; otherwise per-seat limits match Plus for most features.
- API key: billed at API pricing, and model availability follows your key.
Weekly limits may also apply, so check the usage dashboard.
What to do before 14 October
- Search repos,
config.tomlfiles, managed configs and CI scripts forgpt-5.5and replace it. - Admins on Enterprise and Edu should confirm GPT-6 Sol (and Luna if relevant) are enabled in workspace model availability.
- Re-test long agent runs on Sol: per-message allowance (15 to 150) is a wide range and heavy runs sit at the low end.
- If you need GPT-5.5 after the cutoff, move that workload to the API.
Related coverage
- Codex and GPT-6 usage limits by ChatGPT plan
- ChatGPT Pro $100 vs $200 usage explained
- GitHub Copilot retires six models October 19