Not able to disable thinking mode for the self hosted models which use OpenAI as a provider (please read quick description)

[AayushSameerShah]

Before submitting your bug report

• I've tried using the "Ask AI" feature on the Continue docs site to see if the docs have an answer
• I'm not able to find a related conversation on GitHub discussions that reports the same bug
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS: Windows 10
- Continue version: 1.2.11
- IDE version: VS CODE 1.106.3
- Model: Qwen3-8b
- config:
  

C:\Users\aayush\.continu
name: My Config
version: 1.0.0
schema: v1
models:
  - name: Qwen3-8B-chat  
    provider: openai
    model: Qwen/Qwen3-8B
    roles:
      - chat
      - edit
      - autocomplete
    apiKey: RUNPOD_API_KEY
    apiBase: https://api.runpod.ai/v2/XXYYZZ/openai/v1
    defaultCompletionOptions:
      temperature: 0.6
      maxTokens: 4096
      contextLength: 8192
      topP: 0.9
      reasoning: false

    chatOptions:
      baseSystemMessage: |
        Your primary objective is to assist users in their code-writing process through code completion while offering explanations...
      
    requestOptions:
      extraBodyProperties:
        think: false
  
        
    autocompleteOptions:
      temperature: 0.3
      maxTokens: 512
      maxPromptTokens: 2048
      topP: 0.95
      debounceDelay: 200
      onlyMyCode: true
      maxSuffixPercentage: 0.4
      prefixPercentage: 0.6
      useCache: true

    promptTemplates:
      edit: |
        `Here is the code before editing:
        \`\`\`{{{language}}}
        {{{codeToEdit}}}
        \`\`\`

        Here is the edit requested:
        "{{{userInput}}}"

        ## Instructions:
        * DO NOT think
        * Perform edits only.
        * Write the new edited code.

        Here is the code after editing:`
      autocomplete: |
        You are an AI assistant that helps users with code completion. The code context starts with {{{ prefix }}}, ends with {{{ suffix }}}, is part of the file {{{ filename }}}, located in the {{{ reponame }}} repository, and written in {{{ language }}}. Please provide code suggestions that complete the current line or block, considering the surrounding code and the programming language.
context:
  - provider: diff
  - provider: file
  - provider: code

Description

ℹ A little context about my setup

• I have hosted my model on RunPod as a serverless endpoint which used vLLM as backend
• I can totally disable/enable the thinking with the openai interface like shown below ⬇️

chat_response = client.chat.completions.create(
    model="Qwen/Qwen3-8B",
    messages=[
        {"role": "system", "content": system},
        {"role": "user", "content": user},
    ],
    max_tokens=1024,
    temperature=0.6,
    top_p=0.95,
    extra_body={
        "top_k": 20,
        "chat_template_kwargs": {"enable_thinking": False},  # <--- by using this
    },
)

🍒 Problem

As shown in the YAML above, I have tried to disable the thinking in all ways possible:

➡️ Either by:

defaultCompletionOptions:
      temperature: 0.6
      maxTokens: 4096
      contextLength: 8192
      topP: 0.9
      reasoning: false

➡️ Or by:

requestOptions:
      extraBodyProperties:
        think: false

But thinking is being performed no matter what.

I guess, in AutoCompletion it works, but with chat it doesn't.

🤔 Problem with Edit

• When I highlight some code and press CTRL+I to make some inline edits, it uses <think> tags and stops after </think> without writing any actual code.
• When I instruct NOT TO think for edit (as shown in the YAML) it gives empty tags <think> ... </think> and, just that! Nothing else.

🙏🏻 Request

Please fix this, especially in chat mode.

Thanks

[continue]

Related Issues

Found several related issues about thinking/reasoning mode:

Thinking mode control:

• ollama reasoning model won't give think progress #8206 - ollama reasoning model won't give think progress (reasoning disabled by default)
• [Feature Request] Add Thinking Mode Toggle and Filter <think> #5709 - [Feature Request] Add Thinking Mode Toggle and Filter <think> (closed)
• Reasoning models cannot give a thought process #5037 - Reasoning models cannot give a thought process (thinking not displayed for cloud providers)

Provider-specific issues:

• Support for Reasoning in OpenRouter and DeepSeek Providers #5049 - Support for Reasoning in OpenRouter and DeepSeek Providers (different reasoning fields)
• No reasoning with GPT-5 via openAI provider - need support for v1/responses endpoint #7618 - No reasoning with GPT-5 via openAI provider - need support for v1/responses endpoint

Edit mode issues with reasoning:

• BUG: Edit mode with gpt-oss-120b inserts reasoning text into file #8990 - BUG: Edit mode with gpt-oss-120b inserts reasoning text into file
• Reasoning models shows thinking on edit mode after migrating to config yaml. #5822 - Reasoning models shows thinking on edit mode after migrating to config yaml (closed)
• Thinking doesnt works with new LLM Magistral by Mistral! #6278 - Thinking doesnt works with new LLM Magistral by Mistral (closed)

Configuration/format:

• Support configurable thinking output formats #5992 - Support configurable thinking output formats
• self-hosted vllm only works with openai provider but not other openai compatible provider #7051 - self-hosted vllm only works with openai provider but not other openai compatible provider

The core issue appears to be that Continue's handling of reasoning/thinking mode varies significantly across providers, with inconsistent default behavior and limited user control.

[AayushSameerShah]

Update,

• For now I can get away with adding /nothink at the end of my system prompt for chat, but that's not "official" way of doing it and also, in edit mode, I am having issues.

[nicoabie]

I think the problem is that extraBodyProperties is not working.

Relace docs say to send stream: false https://docs.relace.ai/api-reference/instant-apply/apply

This is my configuration:

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: Llama-3.1-8B-Instruct
    provider: openai
    model: meta-llama/Llama-3.1-8B-Instruct
    apiBase: http://localhost:8000/v1
    capabilities:
      - tool_use
    roles:
      - chat
      - edit
      - rerank
      # - autocomplete
      - embed
      - summarize
  - name: Relace Instant Apply
    provider: relace
    model: relace-apply-3
    apiKey: ****
    roles:
      - apply
    requestOptions:
      extraBodyProperties:
        stream: false

This is the error I'm seeing:

[AayushSameerShah]

@nicoabie Damn, why is it so hard !!
Thanks for sharing

[nicoabie]

@nicoabie Damn, why is it so hard !!
Thanks for sharing

Yeah! It was an in other issue I found the solution, check this one out #5644 (comment)