Cmd. /model
Switch among GPT-6 Astra, GPT-5.6 Sol, Terra, Luna, and supported reasoning effort levels.
Key takeaways
- Switch among GPT-6 Astra, GPT-5.6 Sol, Terra, Luna, and supported reasoning effort levels.
- Check the input surface and availability before using the examples below.
Purpose
Switch among GPT-6 Astra, GPT-5.6 Sol, Terra, Luna, and supported reasoning effort levels.
/modelAvailability
| Field | Value |
|---|---|
| Input surface | Running CLI prompt |
| Kind | Built-in |
| Status | Current reference |
| Aliases | — |
| Documented minimum | Not specified |
| Reviewed against | 2026-09-22 · Codex CLI 0.155.1 |
The installed command menu and account or organization settings determine availability.
Examples
/modelWhen to use it
- Select GPT-6 Astra with
high~maxreasoning before the hardest and longest work. - Select Astra or GPT-5.6 Sol before architecture, browser, security, or review work.
- Use Terra for daily implementation and Luna for explicit repeated edits.
- Run
/statusafterward to confirm the model, reasoning, and approval policy.
Choosing related commands
| Command | Difference | Choose It When |
|---|---|---|
/model | Changes model and reasoning | Quality, cost, or latency posture must change |
/fast | Changes speed/service tier | The model is correct but response priority should change |
/status | Shows current policy | You need to verify the selection |
Effects and persistence
Uses the selected model for subsequent requests. Access, reasoning settings, and pricing depend on the chosen model.
/model chooses the active model and, in supported flows, reasoning effort for the current Codex session.
GPT-6 Astra at a Glance
| Item | Value |
|---|---|
| Model ID | gpt-6-astra |
| Minimum CLI | 0.153.0 (bundled default from 0.153.4 when no model is configured) |
| Context / output | 1,050,000 / 128,000 tokens |
| API pricing (per 1M) | $10 input · $1 cached input · $50 output |
| Knowledge cutoff | April 30, 2026 |
| Reasoning | low, medium, high, xhigh, max |
model = "gpt-6-astra"
model_reasoning_effort = "high"The 0.153.4 default change applies only to sessions without an explicit model in config.toml, so
teams that pin a standard should adopt Astra deliberately. Astra also appears in the Amazon Bedrock
Mantle and Runtime pickers from 0.153.3.
Current Selection Notes
- Top model and hardest work: GPT-6 Astra (
gpt-6-astra) - Hardest GPT-5.6 work: GPT-5.6 Sol
- Balanced daily engineering: GPT-5.6 Terra
- Clear repeated or cost-sensitive work: GPT-5.6 Luna
- Pro text-only research preview: GPT-5.3-Codex-Spark
- Previous generation: GPT-5.5 for compatibility or regression comparison
- Retired: GPT-5.4 and GPT-5.4 mini left ChatGPT-sign-in Codex on 2026-08-31
- Remove from new defaults:
gpt-5.2andgpt-5.3-codexfor ChatGPT sign-in
Check permissions before adopting Astra
OpenAI classifies Astra as its first model to reach the Critical cybersecurity capability level
under the Preparedness Framework. Narrow /permissions, sandboxing, and auto-approval scope before
putting it behind broad automation.
Authentication boundary
The August 31 retirement applied to Codex with ChatGPT sign-in. Verify OpenAI API and API-key Codex availability separately in the provider catalog.
Cautions
- Use the account's
/modelpicker for exact IDs and reasoning levels. - Bound concurrency, write scope, and credits before using Ultra delegation.
- Codex Cloud tasks currently use a fixed default model.
- Pair high-risk model choices with
/permissionsand human review.
Official sources
Complete command list · Verification scope · History for this command