No Chain of Thought retrieved if Provider lacks support, despite thinkmode config #4136

Open
opened 2026-02-16 17:42:45 -05:00 by yindo · 7 comments
Owner

Originally created by @DevSissi on GitHub (Jan 3, 2026).

Originally assigned to: @rekram1-node on GitHub.

Description

If the Claude model offered by my model provider does not support the "thinking" capabilities (Thinking Model), does this mean I will be unable to retrieve the Chain of Thought (CoT) even if I have explicitly configured thinkmode in opencode.json?

It seems that the local configuration cannot override the provider's lack of support, effectively resulting in no CoT output.

Plugins

oh-my-opencode @latest, opencode-antigravity-auth @1.2.7

OpenCode version

1.0.223

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

Originally created by @DevSissi on GitHub (Jan 3, 2026). Originally assigned to: @rekram1-node on GitHub. ### Description If the Claude model offered by my model provider does not support the "thinking" capabilities (Thinking Model), does this mean I will be unable to retrieve the Chain of Thought (CoT) even if I have explicitly configured `thinkmode` in `opencode.json`? It seems that the local configuration cannot override the provider's lack of support, effectively resulting in no CoT output. ### Plugins oh-my-opencode @latest, opencode-antigravity-auth @1.2.7 ### OpenCode version 1.0.223 ### Steps to reproduce _No response_ ### Screenshot and/or share link _No response_ ### Operating System _No response_ ### Terminal _No response_
yindo added the bug label 2026-02-16 17:42:45 -05:00
Author
Owner

@github-actions[bot] commented on GitHub (Jan 3, 2026):

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

  • #3284: Claude Sonnet 4.5 is not thinking by default - configuration not being applied
  • #2919: Sonnet 4.5 thinking boxes are not showing despite thinking boxes enabled
  • #3428: Support reasoning configuration for AWS Bedrock provider - related to provider capability limitations
  • #5577: DeepSeek V3.2 reasoning_content not cleared from message history - related to reasoning/thinking handling

These issues involve thinking/reasoning configuration not being properly applied or displayed, which may be related to your situation where thinkmode config is ignored when the provider lacks support. Feel free to ignore if none of these address your specific case.

@github-actions[bot] commented on GitHub (Jan 3, 2026): This issue might be a duplicate of existing issues. Please check: - #3284: Claude Sonnet 4.5 is not thinking by default - configuration not being applied - #2919: Sonnet 4.5 thinking boxes are not showing despite thinking boxes enabled - #3428: Support reasoning configuration for AWS Bedrock provider - related to provider capability limitations - #5577: DeepSeek V3.2 reasoning_content not cleared from message history - related to reasoning/thinking handling These issues involve thinking/reasoning configuration not being properly applied or displayed, which may be related to your situation where `thinkmode` config is ignored when the provider lacks support. Feel free to ignore if none of these address your specific case.
Author
Owner

@rekram1-node commented on GitHub (Jan 3, 2026):

im not sure what u mean whats thinkmode?

@rekram1-node commented on GitHub (Jan 3, 2026): im not sure what u mean whats thinkmode?
Author
Owner

@DevSissi commented on GitHub (Jan 4, 2026):

I am utilizing the Claude models(e.g., Opus 4.5) by kiro(aiclient-2-api). However, it appears that the models in Kiro do not support toggling thinking via the thinking parameter in opencode.json. Is my explanation clear?

@DevSissi commented on GitHub (Jan 4, 2026): I am utilizing the Claude models(e.g., Opus 4.5) by kiro(aiclient-2-api). However, it appears that the models in Kiro do not support toggling `thinking` via the thinking parameter in `opencode.json`. Is my explanation clear?
Author
Owner

@rekram1-node commented on GitHub (Jan 4, 2026):

what is kiro? How are u hitting it? Is this an issue w/ ur plugin? I dont think we maintain a kiro provider

@rekram1-node commented on GitHub (Jan 4, 2026): what is kiro? How are u hitting it? Is this an issue w/ ur plugin? I dont think we maintain a kiro provider
Author
Owner

@DevSissi commented on GitHub (Jan 4, 2026):

In fact, this is the Claude model provided by Amazon. Is there a more convenient way to check if the current model supports "thinking" and simultaneously display whether thinking mode is enabled?
(Kiro is an AI IDE provided by AWS. I used the AIclient-2-API project to convert it into a local API service and integrated it into OpenCode.)

@DevSissi commented on GitHub (Jan 4, 2026): In fact, this is the Claude model provided by Amazon. Is there a more convenient way to check if the current model supports "thinking" and simultaneously display whether thinking mode is enabled? (Kiro is an AI IDE provided by AWS. I used the AIclient-2-API project to convert it into a local API service and integrated it into OpenCode.)
Author
Owner

@rekram1-node commented on GitHub (Jan 4, 2026):

i mean if it is entirely custom model then opencode doesnt know but all the models we maintain lists of we track:

  • reasoning (is it a reasoning model)
  • context limits
  • tool call capabilities
    and much more
@rekram1-node commented on GitHub (Jan 4, 2026): i mean if it is entirely custom model then opencode doesnt know but all the models we maintain lists of we track: - reasoning (is it a reasoning model) - context limits - tool call capabilities and much more
Author
Owner

@DevSissi commented on GitHub (Jan 4, 2026):

By reading Kiro's official documentation, it appears they implement think mode through extensions in their own CLI tool. Does this mean that models from Kiro are inherently models without a think mode?

the doc: https://kiro.dev/docs/cli/experimental/thinking/

@DevSissi commented on GitHub (Jan 4, 2026): By reading Kiro's official documentation, it appears they implement think mode through extensions in their own CLI tool. Does this mean that models from Kiro are inherently models without a think mode? the doc: https://kiro.dev/docs/cli/experimental/thinking/
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#4136