[Feature] Surface reasoning effort controls in chat UI

[jukasper]

Feature Request: UI Controls for Configuring Model Reasoning Levels

Summary
Add in-chat UI controls for configuring the reasoning/thinking effort level for models that support adaptive thinking (e.g. Claude), rather than requiring users to navigate VS Code Settings.

Current Behavior
Reasoning effort can only be configured via VS Code Settings (GitHub > Copilot > Chat > Anthropic > Thinking: Effort), where users must manually set low, medium, or high. This is not discoverable, and community feedback shows many users are unaware this setting exists at all.

Desired Behavior
Expose reasoning level controls directly in the chat UI — for example, via a dropdown, toggle, or model picker — so users can easily switch between effort levels without leaving the chat interface.

User Impact
Multiple users have reported confusion about how to configure this setting:

• Users asking "how do you use xhigh under GitHub Copilot chat?"
• Follow-up confusion about where in VS Code to find the setting

This suggests the current settings-only approach creates unnecessary friction and reduces discoverability.

Proposed Solution
Add a reasoning effort selector to the chat UI, similar to how model selection is surfaced. Options: low, medium, high (matching existing setting values).

References

• Existing setting: GitHub > Copilot > Chat > Anthropic > Thinking: Effort (Preview)
• Related: GitHub > Copilot > Chat: Responses Api Reasoning Effort (Experimental)

[jukasper]

[eli-w-king]

We explored support for configurable reasoning last month, I'm so happy to see this conversation surfaced again here! In a related unmoderated study with 14 users, we found:

• Thinking/reasoning level was understood relatively (low to high), not numerically
• Few users did not expect to configure model reasoning from the model picker

This informed our design explorations here. Happy to take part in this work stream as needed.

[jukasper]

More users input:

[jukasper]

[jukasper]