opencode-go (Console Go) provider returns 400/401/500 for subscribed models #37056

Activity

1. github-actions

   assigned

   MrMushrooooom

   on Jul 15, 2026
2. github-actions commented on Jul 15, 2026

   

   github-actionsBot

   on Jul 15, 2026 – with GitHub Actions

   Contributor

   More actions

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

   • Zen and Go backends fail on DeepSeek V4 / MiMo tool calling with "Internal server error" #36354: Zen and Go backends fail on DeepSeek V4 / MiMo tool calling with "Internal server error" — same providers (opencode-go), same models (deepseek-v4-pro), same upstream error ("Error from provider (Console Go): Upstream request failed")
   • All paid OpenCode Zen models fail with 'Upstream request failed' — free models work #36506: All paid OpenCode Zen models fail with 'Upstream request failed' — overlapping symptom of upstream request failures on paid Go/Zen models
   • GLM-5.2 via OpenCode Go returns 400: messages.content should be a valid string #32821: GLM-5.2 via OpenCode Go returns 400 — same GLM-5.2 model via OpenCode Go returning 400 errors

   If your issue is specifically about the 401 "Missing API key" intermittent rejections or the large-body (300KB+) 400 errors, it may be worth adding that detail to one of the existing reports rather than filing separately.
3. kaka-ruto commented on Jul 16, 2026

   

   kaka-ruto

   on Jul 16, 2026

   More actions

   Having the same issue
4. ctml91 commented on Jul 21, 2026

   

   ctml91

   on Jul 21, 2026

   More actions

   Any solution for this? Or workaround
5. 123456789cm commented on Jul 21, 2026

   

   123456789cm

   on Jul 21, 2026 via email

   Author

   More actions

   I've given up trying to solve this problem; it leaves me helpless.
6. onelpawarai commented on Jul 21, 2026

   

   onelpawarai

   on Jul 21, 2026

   More actions

   Please support my project. https://github.com/onelpawarai/ZYRAXON-AI.git

   👍React with 👍1Reacted by ZYRAXON-AI

   👎React with 👎9Reacted by Alex, lamurian, 油木然yomua, Koresha, adrbn, 王天水, ZYRAXON-AI, 一楼小不点 and Alwyn974

   ❤️React with ❤️1Reacted by ZYRAXON-AI
7. kaka-ruto commented on Jul 21, 2026

   

   kaka-ruto

   on Jul 21, 2026

   More actions

   This workaround worked for me https://x.com/kaka_ruto/status/2077869788152418816
8. panmichalakos-unity commented on Jul 28, 2026

   

   panmichalakos-unity

   on Jul 28, 2026

   More actions

   Same failure here, reproduced from a Letta Code 0.28.17 client on Windows 11, routed through the opencode-go / Console Go provider. Two separate request paths hit it:

   1. Inline parent model: openai Chat Completions API stream failed (400): Error from provider (Console Go): Upstream request failed (run run-72d04346-1209-40b2-b32f-e4b10fa70bec).
   2. Background subagent call — classified llm_bad_request / invalid_request_error, retryable: false (subagent subagent-1785271150477-6, task task_6).

   The Letta server constructs the request payload server-side, so I don't have the exact sampling-parameter values from the failing call. I checked local config (%USERPROFILE%.letta*.json, including the bundled model catalog) and recent *.log files — none of them contain the request body, so I can't confirm whether top_p=0, temperature=0, reasoning_effort, or thinking is the trigger. Reporting those as the cause from my side would be speculation.

   What I can say is that the failure is reproducible from a third-party client (Letta Code) talking to Console Go, and the error is surfaced verbatim — so the 400 is being returned by Console Go / opencode.ai, not invented by the client. This corroborates the umbrella pattern in #37056 and overlaps with the parameter-related report in #37231 (where the underlying Invalid top_p value, the valid range of top_p is (0, 1.0] message was exposed and setting top_p/temperature to null restored functionality).

   Environment: Letta Code 0.28.17, Windows 11, parent model chatgpt-plus-pro/gpt-5.6-terra with reasoning_effort=medium, max_output_tokens=128000, context window 350000.

   Happy to enable debug-level provider logging on the Letta side if that helps isolate the parameter set that Console Go is rejecting.

   +1 to keeping #37056 as the umbrella for this family rather than opening a new report.
9. github-actions commented on Sep 27, 2026

   

   github-actionsBot

   on Sep 27, 2026 – with GitHub Actions

   Contributor

   More actions

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

    closed this as not plannedon Sep 27, 2026