Repository navigation
Zen Go /v1/responses returns 503 for GLM-5.3 and GLM-5.3-Flash #51306
Description
Description
OpenCode Go lists glm-5.3 and glm-5.3-flash, but native Responses clients cannot use either model through POST /zen/go/v1/responses.
On 2026-09-25, requests made from a host running RHEL 10 with the same test-only Go API key returned:
| Model | /v1/chat/completions |
/v1/responses |
|---|---|---|
glm-5.3 |
HTTP 200 | HTTP 503, Endpoint is unavailable |
glm-5.3-flash |
HTTP 200 | HTTP 503, Endpoint is unavailable |
The Responses failure occurs for both streaming and nonstreaming requests. A forced function-tool call with glm-5.3-flash also returns 503. The Chat Completions route returns a valid streaming function call and accepts a follow-up turn with the tool result.
Environment
- Test system: Red Hat Enterprise Linux 10 (RHEL 10)
- Test date: 2026-09-25
- API: deployed OpenCode Go API
- Authentication: test-only Go API key
- The application test suite was run from
packages/console/appon RHEL 10.
Steps to reproduce
- Use a test-only OpenCode Go API key.
- Send requests for
glm-5.3andglm-5.3-flashtoPOST /zen/go/v1/chat/completions; observe HTTP 200. - Send equivalent streaming and nonstreaming requests to
POST /zen/go/v1/responses; observe HTTP 503 withEndpoint is unavailable. - Force a function-tool call with
glm-5.3-flash; observe HTTP 503.
Expected behavior
Both models should return valid Responses JSON or an ordered Responses SSE stream so clients can complete text and multi-turn function-tool turns. The API should preserve session-id and thread-id headers and report usage, including cached tokens when supplied upstream.
Scope and related work
This issue tracks only glm-5.3 and glm-5.3-flash. It is a focused subset of the broader non-DeepSeek outage in #44910. The missing SSE item lifecycle in #40171 is related. Fixing these two GLM models does not by itself resolve either broader issue or the separate Anthropic /v1/messages route.
Proposed fix: #51303.
Activity
github-actions commentedon Sep 25, 2026
This issue doesn't fully meet our contributing guidelines.
What needs to be fixed:
Please edit this issue to address the above within 2 hours, or it will be automatically closed.
This issue might be a duplicate of an existing issue. Please check:
/v1/responsesfailure for non-DeepSeek models, including GLM, while/v1/chat/completionsworks. It reports HTTP 500; this issue narrows the scope to GLM-5.3/Flash and reports HTTP 503.If you believe this was flagged incorrectly, please let a maintainer know.
github-actions commentedon Sep 25, 2026
This issue has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.
Feel free to open a new issue that follows our issue templates.
Validatus commentedon Sep 25, 2026
Hi maintainers — I’ve updated the issue body to include the required
### Descriptionsection and added the test environment (RHEL 10). Could someone with the necessary permissions add theneeds:compliancelabel so the compliance workflow can recheck the edited issue? If it passes, please reopen the issue. The proposed fix is PR #51303 (Fixes #51306). Thank you.