-
Notifications
You must be signed in to change notification settings - Fork 0
Models
ZoyLuo edited this page Jun 7, 2026
·
1 revision
Each connected lane has a model dropdown. It's enabled once the lane connects, and includes a 默认 / Default option (lets the backend pick) plus a refresh button.
| Backend | Source of the list |
|---|---|
| Claude | A curated list mirroring the official Claude Code picker — Opus 4.8, Opus 4.8 (1M context), Sonnet 4.6, Haiku 4.5, plus legacy Opus 4.7 / 4.7 (1M) / 4.6. In API-key mode, any newer models the Anthropic API reports are appended automatically. |
| Codex | Read live from Codex's own cache ~/.codex/models_cache.json (e.g. GPT-5.5 / 5.4 / 5.4-Mini). This updates itself as Codex refreshes the cache. |
| DeepSeek | Fetched live from the DeepSeek /models API (falls back to deepseek-chat / deepseek-reasoner). |
The model id you pick is passed straight to the CLI (claude --model …, codex -m …) or API. The [1m] suffix on Claude selects the 1M-context variant.
If a backend has no fetchable list (e.g. Codex when its cache is missing), the picker falls back to a free-text field so you can type any model id. When a backend does have a list but you've set a model that isn't in it, your custom value still shows as the selected option.
Hit the ⟳ button next to the picker to re-fetch the list — handy right after adding an API key or when a provider ships a new model.