Run GPT-5.6 in Codex CLI
As of 2026-11-10 there are three ways to run GPT-5.6 in Codex CLI: launch with codex -m gpt-5.6-sol, put model = "gpt-5.6-sol" in ~/.codex/config.toml, or pick it with /model inside a session (Terra and Luna switch the same way); the current stable Codex CLI is 0.167.0 (2026-11-06). The live question is not “did 5.6 ship” but what to do when Sol is missing from /model: the official docs say /model only shows models available to your account and workspace, and with an API key, available models follow your API organization and project; ChatGPT-signed-in Codex has not offered GPT-5.4 since 2026-08-31. With a QCode provider, Codex uses QCode's base_url and key, and you pick gpt-5.6-sol or gpt-5.6-terra, or a GPT-6 model from /models.
Why doesn't /model show gpt-5.6?
Two model lists get mixed up: ChatGPT-signed-in Codex is not the same list as an API-key / QCode provider. The official docs say /model only shows models available to your account and workspace; as of 2026-11-10 the recommended list on the official Codex models page is gpt-6-astra, gpt-6.1-sol, gpt-6.1-terra and gpt-6-luna, and GPT-5.6 is not on it. Older CLI builds also omit the three gpt-5.6-* tiers from /model. Confirm how you authenticate, then update the CLI; if the row is still missing, set the model directly with -m gpt-5.6-sol or model = "gpt-5.6-sol" — do not treat a missing row as a retirement first.
Before you start
- Node.js v22 LTS or later and npm (to install / update Codex CLI with npm, per the QCode docs)
- A QCode.cc API key (issued when you activate any plan, from $9/mo; one key covers Claude / GPT / Chinese models)
- About 5 minutes
Step 1: Update Codex CLI to the latest version
The GPT-5.6 tiers ship built in from releases around July 2026: in the official changelog, Codex CLI 0.144.6 (2026-07-18) lists “Refreshed bundled instructions for GPT-5.6 Sol, Terra, and Luna”. As of 2026-11-10 the current stable release is 0.167.0 (2026-11-06). One global update command does it: npm install -g @openai/codex@latest; confirm with codex --version afterwards.
# update to the latest Codex CLI
npm install -g @openai/codex@latest
# confirm the version
codex --version
Step 2: Point the Codex provider at QCode in config.toml
Point the provider at QCode in ~/.codex/config.toml: model_provider = "qcode", then under [model_providers.qcode] set base_url = "https://api.qcode.cc/openai", wire_api = "responses" and requires_openai_auth = true; the API key goes in OPENAI_API_KEY inside ~/.codex/auth.json (with requires_openai_auth = true, Codex reads it from there). OpenAI's docs say current Codex releases no longer support wire_api = "chat", so a custom provider needs the Responses interface. Full details are in the Codex tutorial in the QCode docs.
# ~/.codex/config.toml — point the provider at QCode
# full config reference:
# https://docs.qcode.cc/en/docs/getting-started/codex-quick-start
model = "gpt-5.6-terra" # default tier; sol also available
# ... keep your QCode provider / auth settings unchanged
Step 3: Pick a tier with /model (or set it in config)
Launch codex and type /model to pick a model, or write /model gpt-5.6-sol directly (since Codex CLI 0.164.0, /model accepts arguments); codex -m gpt-5.6-sol at launch works too, and model = "gpt-5.6-sol" in config.toml sets the default. The models on sale at QCode are listed on /models — they include gpt-5.6-sol, gpt-5.6-terra, gpt-6.1-sol, gpt-6.1-terra, gpt-6-sol and gpt-6-astra; For light tasks on QCode, use gpt-5.6-terra (QCode does not currently support gpt-5.6-luna or gpt-6-luna). Rule of thumb: Terra as default, Sol for the hard parts.
$ codex -m gpt-5.6-terra
> /model
gpt-5.6-sol
● gpt-5.6-terra ← selected
Step 4: Verify it runs and bills correctly
Run a small task to confirm responses, and run /status in the session to confirm the active model; then check your QCode dashboard usage — you should see calls under the gpt-5.6 tier you picked. All tiers share the same quota as the other models on QCode; nothing extra to buy.
$ codex "add a unit test for utils/retry.ts"
# response streams from gpt-5.6-terra
# then: qcode.cc dashboard → usage
# calls appear under the gpt-5.6 tier you selected
FAQ
I can't update the CLI (version-pinned). Can I still use GPT-5.6?
Yes. Older versions can register models manually via the model-catalog mechanism — the process is identical to our earlier model-catalog enablement guide, just with gpt-5.6-sol / terra as the model ids in the JSON. Updating the CLI remains the preferred path though: one command replaces all manual steps.
Which tier should I pick?
Terra for daily work; Sol for hard jobs like cross-repo refactors; for completions, batch and light CI tasks — OpenAI's Luna tier — use gpt-5.6-terra on QCode. If you can move to the GPT-6 generation: OpenAI's changelog (2026-09-22) says GPT-6 Sol and GPT-6 Luna come at lower token prices than their GPT-5.6 predecessors, and as of 2026-11-10 the official Codex models page describes GPT-6.1 Sol as near-Astra performance at a lower cost than Astra; gpt-6.1-sol and gpt-6-sol are available on QCode too. See our Sol vs Terra vs Luna guide for the full comparison. Sol vs Terra vs Luna →
How do billing and quota work?
gpt-5.6-sol and gpt-5.6-terra share one QCode key and one quota with the other GPT models on QCode, Claude and Chinese models, billed per token; per-model rates are on /models (as of 2026-11-10 the GPT models on sale are gpt-6.1-sol, gpt-6.1-terra, gpt-6-sol, gpt-6-astra, gpt-5.6-sol and gpt-5.6-terra). Not ready for a monthly plan? Pay-as-you-go Packs give you a one-time quota to use within 7 days (see /pay-as-you-go); if you already have a Personal Plan, add a booster pack when you need more quota.
Any difference between calling via QCode and OpenAI directly?
The interface is the same standard one: Codex connects to QCode over the OpenAI Responses protocol, with base_url pointing to https://api.qcode.cc/openai, wire_api set to "responses", and the key in ~/.codex/auth.json. The difference is the onramp: QCode gives you one key across Claude / GPT / Chinese models, the same key also works in Claude Code, and switching models needs no extra setup; billing is per token, with per-model rates on /models.
Try GPT-5.6 in Codex now
Create a QCode account, activate a plan from $9/mo to get your API key, and run gpt-5.6-sol / gpt-5.6-terra in Codex CLI within five minutes.
Create a QCode account
One key across Claude, GPT and Chinese models, billed per token — live rates on /models.