Issue Station
Sign In
Tools hereToolbox

gemini-cli - 💡(How to fix) Fix Interactive Thinking Budget / Reasoning Effort Control (Competitive Parity) [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
google-gemini/gemini-cli#25122•Fetched 2026-04-11 06:31:23
View on GitHub
Comments
1
Participants
2
Timeline
5
Reactions
0
Author
nick-youngblutnick-youngblut
Participants
gemini-cli[bot]gemini-cli[bot]nick-youngblutnick-youngblut
Timeline (top)
labeled ×3commented ×1issue_type_added ×1
RAW_BUFFERClick to expand / collapse

What would you like to be added?

An interactive mechanism to adjust the model's thinking budget or reasoning effort during a session, without requiring manual edits to settings.json. This could take the form of:

  1. Slash command: /think [off|low|medium|high|max] or /budget <tokens> to adjust thinkingBudget (2.5 models) or thinkingLevel (Gemini 3 models) for subsequent prompts.
  2. Natural language keywords: Detect phrases like "think hard about this" or "think deeply" in the prompt and automatically increase the thinking budget for that request.
  3. Sticky toggle: A /think toggle that persists for the remainder of the session, with a visual indicator in the UI showing the current thinking level.

Why is this needed?

Every major competing CLI coding agent already supports this, and Gemini CLI's lack of it is a concrete disadvantage in complex engineering tasks.

Competitive landscape

ToolInteractive thinking controlMechanism
Claude Code✅Natural language keywords: "think", "think hard", "ultrathink" scale extended thinking from normal → high → maximum budget. No config file edits needed.
OpenAI Codex CLI✅reasoning_effort parameter (low, medium, high) configurable per session.
Gemini CLI❌Requires manually editing settings.json or modelAliasOverrides. No per-session or per-prompt control. Changes require restarting the CLI.

This gap matters in practice. When working on complex tasks — system architecture, multi-file refactors, debugging subtle race conditions — the ability to tell the agent "think harder about this one" in the moment is essential. Having to stop, edit a JSON file, and restart the CLI to get deeper reasoning breaks flow and is simply not competitive.

Additional context

Real-world impact

Users doing quick tasks (renaming variables, generating boilerplate) want fast, low-budget responses. Users debugging a complex distributed systems issue in the same session want maximum reasoning depth. Today, you can't have both without leaving the CLI. Claude Code handles this seamlessly — you just type "ultrathink" in your prompt and get maximum extended thinking for that response, then the next quick question goes back to normal.

This is consistently cited in community comparisons as a reason Gemini CLI underperforms on hard tasks (see #6693), and is likely a significant factor in perceived quality gaps rather than the underlying model capability itself.

extent analysis

TL;DR

Implementing an interactive mechanism to adjust the model's thinking budget during a session would bridge the gap with competing CLI coding agents.

Guidance

  • Introduce a slash command, such as /think [off|low|medium|high|max], to adjust the thinking budget for subsequent prompts.
  • Develop natural language keyword detection to automatically increase the thinking budget based on phrases like "think hard about this".
  • Create a sticky toggle with a visual indicator in the UI to show the current thinking level, allowing users to persistently adjust the thinking budget for the remainder of the session.
  • Consider implementing a reasoning_effort parameter, similar to OpenAI Codex CLI, to provide configurable per-session control.
  • Evaluate the feasibility of integrating these mechanisms without requiring manual edits to settings.json or restarting the CLI.

Example

/think high

This could be an example of a slash command to increase the thinking budget for subsequent prompts.

Notes

The implementation details may vary depending on the underlying architecture and model capabilities of Gemini CLI. It is essential to ensure that the introduced mechanisms do not compromise the performance or accuracy of the model.

Recommendation

Apply a workaround by introducing a slash command or natural language keyword detection to adjust the thinking budget, as this would provide an immediate solution to bridge the gap with competing CLI coding agents.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#database connection#vector store#embedding generation#cache error#pipeline error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING