docs: document thinking variants for sandbox models (#26292)

Document how to select a model reasoning level with `sbx run --variant`,
with an example and the relevant agent limitations.

Validation: Markdown and style checks pass.

Generated by Codex.
This commit is contained in:
David Karlsson authored and GitHub committed 2026-10-13 15:11:14 +02:00
1 parent 89f0a32182
commit 734ce2eeb8
1 file changed
+19
@@ -186,6 +186,25 @@ Changing the model recreates the sandbox container. The workspace and
kit-owned volumes persist. Omit `--provider` to select a local model managed
by `llmman`.
## Choose a reasoning level
Use `--variant` with `--model` to select a reasoning level supported by your
model. For example, replace `<MODEL_NAME>` with an OpenAI model that supports
`high` reasoning effort to start Codex at that level:
```console
$ sbx run --provider openai --model <MODEL_NAME> --variant high codex
```
Available values depend on the model and agent. Reasoning levels are `minimal`,
`low`, `medium`, `high`, `xhigh`, and `max`. Some models support `none` to disable
thinking or `thinking` to enable it. Omit `--variant` to use the default.
Claude Code doesn't support `none` or `minimal`. For Claude Code and Codex,
`--variant` requires an attached session and can't be combined with `--detach`.
In OpenCode, it selects the initial variant, which you can change inside
OpenCode.
## Use another provider for larger requests
Pair a local model with another provider to handle requests that exceed the