OpenAI · Models
Codex Models Now, and What Replaces GPT-5.5
GPT-5.5 retires from Codex with ChatGPT sign-in on October 14, 2026. Switch to GPT-6 Sol (gpt-6-sol) on Plus, Pro, Business, Enterprise and Edu, or GPT-6 Luna (gpt-6-luna) on Free and Go. For new complex work OpenAI recommends GPT-6.1 Sol where your account has it. Codex used with an API key is not affected.
Direct answer
Which Codex model should I use now?
- GPT-5.5 deadline
- Fact GPT-5.5 retires from ChatGPT, ChatGPT Work and Codex on all plans on October 14, 2026; the OpenAI API is not affected. GPT-5.3-Codex-Spark retired on September 14, 2026. GPT-5.4 and GPT-5.4 mini retired from ChatGPT-signed-in Codex on August 31, 2026.
- Replacement for GPT-5.5
- Fact GPT-6 Sol on Plus, Pro, Business, Enterprise and Edu. GPT-6 Luna in the desktop app on Free and Go. Both are subject to rollout.
- Recommended default
- Fact GPT-6.1 Sol is the recommended model for complex coding where available; GPT-6 Luna is recommended for focused, repeatable tasks. GPT-5.6 Sol, Terra and Luna remain available during the rollout.
- Free and Go
- Fact GPT-6 Luna only. GPT-6.1 Sol is not included on Free and Go at launch.
- Last verified
- Oct 11, 2026, 9:30 AM UTC
Codex models by plan
Fact Availability depends on rollout, sign-in method and client. Selecting a model does not change workspace permissions or grant access to it.
| Model | Config id | Use it for | Free | Go | Plus | Pro | Business | Plus / 5h |
|---|---|---|---|---|---|---|---|---|
| GPT-6 Luna | gpt-6-luna | Most efficient model for focused, repeatable, high-volume tasks. GPT-5.5 replacement on Free and Go. | Yes | Yes | Yes | Yes | Yes | 350–3,000 |
| GPT-6.1 Sol | gpt-6.1-sol | Recommended for complex coding and long-running work. Near-Astra performance at a lower cost than Astra. | No | No | Yes | Yes | Yes | 15–160 |
| GPT-6 Astra | — | Most capable model, for the hardest end-to-end work across code, apps and research. | No | No | Yes | Yes | Yes | 5–45 |
| GPT-6 Sol | gpt-6-sol | Official GPT-5.5 replacement on Plus, Pro, Business, Enterprise and Edu. | No | No | Yes | Yes | Yes | 15–150 |
| GPT-5.6 Sol | — | Previous generation. Remains available during the GPT-6 rollout. | No | No | Yes | Yes | Yes | 10–100 |
| GPT-5.6 Terra | — | Previous generation. Remains available during the GPT-6 rollout. | No | No | Yes | Yes | Yes | Not in the Plus table |
| GPT-5.6 Luna | — | Previous generation. Remains available during the GPT-6 rollout. | No | No | Yes | Yes | Yes | 250–2,000 |
Plan prices: Free $0.00/month · Go $8.00/month · Plus $20.00/month · Pro $100.00/month · Business $20.00/user/month, billed annually, minimum 2 seats. Pro plans currently have no five-hour limit, so Pro has no range. Full limits are on Codex usage limits; plan costs are on Codex pricing.
Retired and retiring Codex models
| Model | Id | Status | Date | Official replacement |
|---|---|---|---|---|
| GPT-5.5 | gpt-5.5 | Retiring | October 14, 2026 | gpt-6-sol on Plus, Pro, Business, Enterprise and Edu; gpt-6-luna on Free and Go |
| GPT-5.3-Codex-Spark | gpt-5.3-codex-spark | Retired | September 14, 2026 | Any model available to your account and client |
| GPT-5.4 | gpt-5.4 | Retired | August 31, 2026 | gpt-6-sol |
| GPT-5.4 mini | gpt-5.4-mini | Retired | August 31, 2026 | gpt-6-luna (Enterprise and Edu: an administrator must enable Luna first) |
GPT-5.2 and GPT-5.3-Codex were already deprecated in Codex with ChatGPT sign-in.
Migration checklist
How to switch off GPT-5.5
- Default model. The ChatGPT desktop app, Codex CLI and IDE extension share
config.toml. Replacemodel = "gpt-5.5"withmodel = "gpt-6-sol"(or"gpt-6-luna"on Free and Go). With no model entry, Codex uses a recommended model. - CLI sessions. Use
/modelin an interactive session, or launch withcodex --model gpt-6-sol. Non-interactive runs take the same flag:codex exec -m gpt-6-sol "…". - Everything else that names a model. Workspace defaults, saved model settings, managed configurations, custom agents, scheduled tasks and scripts.
- Check the limit, not just the name. A newer model can use your five-hour allowance at a different rate. Compare the Plus ranges above, and see “Selected model is at capacity” if a model refuses requests while usage remains.
Unknown When the GPT-6.1 Sol rollout reaches every account and client. If it is not in your model picker, GPT-6 Sol is the documented fallback.
Source: Codex models and the Codex usage-limit table.