In the Codex App or CLI, Sol is suited to routine code reviews and making code changes based on review feedback. For complex cross-service troubleshooting, choose Astra first.
Suitable tasks: Review a pull request and identify issues. Modify code based on review feedback and run the relevant tests.
Unsuitable tasks: Trace a bug across services, fix it, and verify the affected workflow. The source lists this type of task as a use case for Astra.
Applicable model version: GPT-6.2 Sol.
Applicable client, agent, or API: Codex App, Codex CLI. This article records model selection in Codex only and does not cover Responses API settings.
Recommended reasoning level and parameters: The source provides no Codex-specific reasoning level or parameter recommendations.
The source does not provide a complete, copyable prompt. Based on its description, you can frame the task as follows: ask Sol to review a PR and list issues that need to be addressed. Or provide review feedback and ask Sol to modify the implementation and run the relevant tests. This is a Chinese paraphrase of the source's task descriptions.
Select Sol in the model picker in the Codex App or by using /model in the Codex CLI.
Choose a short task: review a PR, or modify code based on existing review feedback.
For modification tasks, ask Sol to run the relevant tests. Switch to Astra for complex cross-service diagnosis.
The page says that models can be selected in the Codex App and CLI through the model picker or /model.
The page lists “review a pull request and flag issues for the author to address” and “modify code based on review feedback and run the relevant tests” as Sol use cases.
The page lists tracing a bug across services, fixing it, and testing the affected workflow as an Astra use case.
The page provides no benchmark data, test counts, or outputs for these Codex workflows, and does not specify a Codex reasoning level.
This guidance supports model and task-type selection, as well as where to select the model. It does not specify prompt formats, how to organize codebase context, test commands, or acceptance criteria. Those details depend on the repository and task. The default reasoning level mentioned for Work should not be applied directly as a Codex setting.
Open the model selector in the Codex App or CLI, select GPT-6.2 Sol, and submit a PR review task, or provide existing review feedback and ask Sol to make changes and run the relevant tests. Record the selected model, task input, modification results, and test results. The source does not define a standard format for reproduction data.
GPT-6.2 Sol