Error: using organization model. 400

[fzhang00]

I'm using Azure hosted gpt-5.1 model from my organization.

Continue installation is through extention in VSCode, Windows11.

The config yaml is setup like this:

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: "myorg-Global-gpt-5.1-chat"
    provider: "openai"
    model: "myorg-Global-gpt-5.1-chat"
    apiBase: 'https://server-address.azure.com/openai/v1/'
    apiVersion: "2025-11-13"
    apiKey: 'my_api'

The setup would work sometimes, but in the same session when it is working, the next conversation can produce this error below. Usually if I use the resubmit last message, it usually work, but submitting new message still not working. Please advice how to fix this issue. Thanks.

Error Details

Model: myorg-Global-gpt-5.1-chat
Provider: openai
Status Code: 400

Error Output

{"message":"Item 'rs_01dffa94d3f99bb000694ab5f2528c8195b97657fa78ccba22' of type 'reasoning' was provided without its required following item.","type":"invalid_request_error","param":"input","code":null}

Additional Context
Please add any additional context about the error here

[continue]

Related issues about reasoning tokens and 400 errors:

• 'reasoning' was provided without its required following item #8933 - Same error: 'reasoning' was provided without its required following item
• Error: G 5 - 400 #9264 - GPT-5 400 error with reasoning token issue
• Error: GPT-5 - 400 #9155 - GPT-5 400 error with tool call issues
• Error: GPT-5 - 400 #9164 - GPT-5 400 error with organization verification for reasoning summaries
• Bug: DeepSeek R1 (Reasoner) fails with MCP tools - 400 Missing reasoning_content #9245 - DeepSeek R1 fails with MCP tools - 400 Missing reasoning_content
• When use deepseek v3.2 as agent, throw "400 Missing reasoning_content field in the assistant message at message index 2" #8989 - DeepSeek v3.2 as agent throws "400 Missing reasoning_content field"

These issues appear to be related to how Continue handles reasoning tokens/content in the conversation history when using models with reasoning capabilities (GPT-5, DeepSeek R1/v3.2, etc.), especially in multi-turn conversations with tool calling.