Have an ability to turn on / off thinking mode

Better yet, customisable based on models.

e.g:

GPT-5 we can toggle between minimal / low / medium / high
Sonnet 4.5, we can toggle between non thinking and thinking

This issue might be a duplicate of existing issues. Please check:

Support for reasoning_effort parameter in UI #450: Requests UI controls for reasoning_effort parameter, similar to thinking mode toggles
请求添加thinking UI #1177: Specifically requests thinking UI features and controls
[Feature Request] Add support for using thinking tags in local models #1528: Requests better support for thinking tags/mode in local models
Better to optimize the agent ouput by distinguishing the thinking process and the final response #2767: Discusses optimizing thinking process output and presentation

Feel free to ignore if none of these address your specific case.

Absolutely makes a lot of sense, I think Dax & Adam have talked about UX of this and once opentui migration is complete this will be added at some point

Also came here because I'm really missing this. The ability to toggle thinking on/off especially for Sonnet is really handy!

I found the config that enables it, but it's permanent and being able to decide for the task at hand would be good

I found the config that enables it, but it's permanent and being able to decide for the task at hand would be good

Where is that option? Usually you can write an agent with a particular config, so you might write @think something to use a thinking-enabled agent.

Will you make thinking on by default for thinking models?
unless the toggle for thinking mode is super obvious, new users coming from claude might not realize they need to toggle thinking

I found the config that enables it, but it's permanent and being able to decide for the task at hand would be good

Where is that option? Usually you can write an agent with a particular config, so you might write @think something to use a thinking-enabled agent.

^^ also like this idea, to trigger thinking explicitly from a prompt allows you to activate it for that 1 prompt if you dont want thinking enabled by default.

I have this in my opencode.jsonc, so toggling thinking is just changing the model, but still, a thinking toggle would be ideal:

  "provider": {
    "anthropic": {
      "models": {
        "[Custom] Opus 4.5 Latest / Thinking": {
          "id": "claude-opus-4-5",
          "options": {
            "thinking": {
              "type": "enabled",
              "budgetTokens": 16000
            }
          }
        },
        "[Custom] Sonnet 4.5 Latest / Thinking": {
          "id": "claude-sonnet-4-5",
          "options": {
            "thinking": {
              "type": "enabled",
              "budgetTokens": 16000
            }
          }
        }
      }
    }
  },

coming soon!

I also think that triggering it at prompt level would be really cool

Navigation Menu

Repository navigation

Have an ability to turn on / off thinking mode #2875

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Have an ability to turn on / off thinking mode #2875

Description

Activity

github-actions commented on Sep 30, 2025

rekram1-node commented on Sep 30, 2025

markus1189 commented on Nov 22, 2025

llucax commented on Nov 27, 2025

bcheung commented on Dec 9, 2025

jasonish commented on Dec 9, 2025

rekram1-node commented on Dec 9, 2025

franroa commented on Dec 23, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions