Error: GPT-5.4 - 400

[MoussaSak]

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 11
- Continue version: 1.2.22
- IDE version: VsCode 1.123.0
- Model: GPT 5.4 on Azure
- config:
  
  - name: GPT-5.4
    provider: openrouter
    model: azure/gpt-5.4
    apiBase: http://localhost:3130
    roles:
      - chat
      - edit
      - apply
      - autocomplete
    capabilities:
      - tool_use

Description

Am using LiteLLM as local AI Gateway to consume some models like gpt-5.4 on an Azure, and it seems that continue extension is having some problems with autocomplete function on the recent OpenAI models (i've tested gpt-5.4-mini also).

To reproduce

Just try the same model behind an AI gateway like LiteLLM and you will get the same message

Log output

`Continue Autocomplete Error: HTTP 400 Bad Request from http://127.0.0.1:3130/completions {"error":{"message":"litellm.UnsupportedParamsError: azure does not support parameters: ['stop'], for model=azure/gpt-5.4. To drop these, set `litellm.drop_params=True` or for proxy:\n\n`litellm_settings:\n drop_params: true`\n. \n If you want to use these params dynamically send allowed_openai_params=['stop'] in your request.. Received Model Group=azure/gpt-5.4\nAvailable Model Group Fallbacks=None","type":"None","param":null,"code":"400"}} `

[HerShin5]

This looks like the gateway is forwarding stop into an Azure model path where it isn’t accepted, and autocomplete hits /completions, so it shows up there first.

Two things I’d try:

1. In LiteLLM proxy config, make sure drop_params: true is under litellm_settings and restart the proxy, not just the Continue extension.
2. If it still happens, split chat/edit and autocomplete into separate model entries, and use a model/gateway route for autocomplete that normalizes unsupported OpenAI params.

I’ve run into the same class of issue when testing OpenAI-compatible gateways. With a Base URL like https://kiroforge.cloud/v1, the important bit is not just “OpenAI-compatible” but whether the gateway/provider strips or maps unsupported params consistently, otherwise Continue autocomplete breaks before chat does.

[github-actions]

This issue hasn't been updated in 90 days and will be closed after an additional 10 days without activity. If it's still important, please leave a comment and share any new information that would help us address the issue.

[github-actions]

This issue was closed because it wasn't updated for 10 days after being marked stale. If it's still important, you are welcome to leave a comment and this will keep the issue open. However please note that this repo is no longer being actively maintained.

[michaelben2336-ctrl]

What happnd

[michaelben2336-ctrl]

How do I find my mit license

[michaelben2336-ctrl]

How do I get and use software decoder licenses

[michaelben2336-ctrl]

How do get amd use softwear decoder

…

On Mon, Sep 14, 2026, 8:49 PM github-actions[bot] ***@***.***> wrote: *github-actions[bot]* left a comment (continuedev/continue#12560) <#12560 (comment)> This issue was closed because it wasn't updated for 10 days after being marked stale. If it's still important, you are welcome to leave a comment and this will keep the issue open. However please note that this repo is no longer being actively maintained. — Reply to this email directly, view it on GitHub <#12560?email_source=notifications&email_token=CMJYUZCSD77HZGJD2SAUBPT5PCNZPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRXGM2DGOBXHAZKM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-5673438782>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CMJYUZCWEEHYXYDMWTNMQML5PCNZPAVCNFSNUABFKJSXA33TNF2G64TZHM3DINBWHA3DSMBVHNEXG43VMU5TINJZGU4TEMJTGI32C5QC> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>