qwen3.7-max returns 401 unsupported_value for response_format.type via oa-compat

Bug Description

Using qwen3.7-max through OpenCode Go (oa-compat endpoint) returns a 401 error related to response_format.type.

error.code=unsupported_value_401
error.param=response_format.type
openai.RequestError: Response status code indicates failure

Other models (e.g. mimo-v2-pro, kimi-k2.6) work fine with the same config.

Related Issues

401 not supported for format oa-compat #29568 — same error, reported May 27
Bug: qwen3.7-max returns ModelError "not supported for format oa-compat" via Zen/Go #29688 — marked as duplicate of 401 not supported for format oa-compat #29568
[BUG]: qwen3.7-max fails from Claude Code with Alibaba 401 response #29558 — qwen3.7-max also failing from Claude Code

Steps to Reproduce

Configure OpenCode Go provider with qwen3.7-max model
Make any request (including a simple "hi")
Observe 401 error on response

Expected Behavior

Model should respond normally like other models do.

Actual Behavior

Immediate 401 error:

error.code=unsupported_value_401
error.param=response_format.type

Environment

Model: qwen3.7-max
Provider: opencode-go (oa-compat)
Other models (mimo-v2-pro, kimi-k2.6) work fine

Suspected Cause

The oa-compat layer may be sending a response_format parameter that qwen3.7-max doesn't support (e.g. structured output / JSON mode). The model's API rejects it as unsupported.

Additional Context

Multiple users reporting this in the past few days. Seems to be a regression or incompatibility introduced with the qwen3.7-max model update.

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

401 not supported for format oa-compat #29568: Same 401 unsupported_value error for response_format.type via oa-compat
Bug: qwen3.7-max returns ModelError "not supported for format oa-compat" via Zen/Go #29688: Bug report for qwen3.7-max returning ModelError "not supported for format oa-compat" via Zen/Go (marked as duplicate of 401 not supported for format oa-compat #29568)

Please follow the discussion in #29568 for updates and consider closing this issue if it covers the same problem.

May I ask what you are suggesting with this screenshot? The API endpoint for Opencode needs to be changed to use Qwen 3.7, maybe? So, you suggest, I need one endpoint and protocol for Qwen3.7 and others for the other models? Please clarify. This could maybe be solved with a litelllm server, if I understand your screenshot correctly. But then Opencode shouldn't report Qwen 3.7 on the default endpoint IMO. Kind regards

@Cephei-OpenSource

@Cephei-OpenSource Yes, based on tests from others online, the configuration for Qwen 3.7 differs from that of other models. Other models use the OpenAI Chat Completions API, but Qwen 3.7 requires the Anthropic Messages protocol to function properly.

try this:

curl https://opencode.ai/zen/go/v1/messages \  
-H "Content-Type: application/json" \  
-H "x-api-key: $OPENCODE_GO_API_KEY" \  
-H "anthropic-version: 2023-06-01" \  
-d '{"model":"qwen3.7-max","messages":[{"role":"user","content":"hi"}],"max_tokens":20}'`

https://mp.weixin.qq.com/s/PIMju3QjgBOzhtgXyTSZ_A

Thanks a lot. I can confirm with the anthropic protocol Qwen 3.7 Max is working with Open WebUI if embedded via a LiteLLM Server to force the different protocols. In case somebody needs it:

Declare the API Key for Opencode in .env
Don't forget to define a variable for it also in docker-compose.yml
Entries for config.yaml (example models: MiMo V2.5 Pro and Qwen 3.7 Max):

 # Opencode
  # 1. Mimo -> OpenAI Format, base needs /v1, protocol is openai/
  - model_name: mimo-v2.5-pro
    litellm_params:
      model: openai/mimo-v2.5-pro
      api_base: https://opencode.ai/zen/go/v1
      api_key: os.environ/OPENCODE_API_KEY
      additional_drop_params: ["reasoningSummary"]

  # 2. Qwen -> Anthropic Format, base without /v1, protocol is anthropic/
  - model_name: qwen3.7-max
    litellm_params:
      model: anthropic/qwen3.7-max
      api_base: https://opencode.ai/zen/go
      api_key: os.environ/OPENCODE_API_KEY

This example works for a Zen subscription using Go models.

Again, thanks a lot for your help. Kind regards, Michael

still waiting for this, lol

Might be related to auth/config errors from OpenAI. npx ai-doctor might catch API key issues and auth config problems.

To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.

Navigation Menu

Repository navigation

qwen3.7-max returns 401 unsupported_value for response_format.type via oa-compat #29754

Bug Description

Related Issues

Steps to Reproduce

Expected Behavior

Actual Behavior

Environment

Suspected Cause

Additional Context

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

qwen3.7-max returns 401 unsupported_value for response_format.type via oa-compat #29754

Description

Bug Description

Related Issues

Steps to Reproduce

Expected Behavior

Actual Behavior

Environment

Suspected Cause

Additional Context

Activity

github-actions commented on May 28, 2026

CJ5035 commented on May 29, 2026

Cephei-OpenSource commented on May 29, 2026

CJ5035 commented on May 30, 2026

Cephei-OpenSource commented on May 30, 2026

moinulmoin commented on May 31, 2026

carter-smalls commented on May 31, 2026

github-actions commented on Jul 31, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions