Repository navigation
Internal Server Error #35486
Description
Description
Even with a new session, clearing cache, I get "internal server error" with DeepSeek v4 Flash.
Plugins
No response
OpenCode version
1.17.13
Steps to reproduce
- Launch OpenCode Go in a window with Admin privs.
- Connect to opencode with known good key and subscription.
- Start a new session even with a basic "hello".
- Fails with "Internal Server Error."
Screenshot and/or share link
Operating System
Windows 11
Terminal
Windows Terminal
Metadata
Metadata
Assignees
Labels
No labels
Activity
github-actions commentedon Jul 5, 2026
This issue might be a duplicate of existing issues. Please check:
Please review these issues to see if your problem is already tracked there.
opencodevibrant commentedon Jul 5, 2026
Same issue for me on DeepSeek V4 Flash and MiMo-V2.5. Any new chat and some existing chats shows internal server error.
asrone-builder commentedon Jul 5, 2026
same issue, fro mimo 2.5, and deepseek v4
rustykris commentedon Jul 5, 2026
same here, for DeepSeek V4 Flash + MiMo-V2.5
tested MiniMax M2.7, Qwen3.6 Plus and these responded correctly
willoak commentedon Jul 6, 2026
No provider available
kimi k2.7 code
aadielpr commentedon Jul 6, 2026
same here, mimo 2.5 and deepseek from opencode-go provider, shows internal server error. i try the other providers, and its working.
Device: Mac OS
Opencode (TUI) version: 1.17.13
Plugin: Nothing
Server: Singapore
hrodriguez72 commentedon Jul 6, 2026
Same thing using opencode Go suscription from Opencode CLI 1.17.13 affecting all models. Same message: API call failed after 3 retries: HTTP 500: Internal server error".
dattarohu-coder commentedon Jul 6, 2026
same, with deepseek 4 flash
fahadabdulaziz commentedon Jul 6, 2026
same here, I noticed:
DeepSeek v4 Flash/Pro = Internal Server error
DeepSeek v4 Flash Free = working !
asimons81 commentedon Jul 6, 2026
Confirmed via direct API testing of /zen/go/v1 — this is model-specific, not a general Go outage.
Models returning 500:
Models working (same key, same endpoint):
Quick repro:
curl -s -X POST https://opencode.ai/zen/go/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $KEY" -d '{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"hi"}],"max_tokens":5}'
Returns 500. Same call with qwen3.5-plus returns a valid completion. Models endpoint lists them all fine — the inference backends for DeepSeek/MiniMax are the problem.
yaephin commentedon Jul 6, 2026
这也太烂了吧，一个月断几次，服务器太垃圾了
asimons81 commentedon Jul 6, 2026
Seems to be back up and working now.
PawelGIX commentedon Jul 27, 2026
Im getting AI_APICallError: Internal server error on big-pickle for week now.
Sometimes it works sometimes it doesn't work.
Lately it doesn't work more often.
Does anyone else experience these types of problems?
Screen.Recording.2026-07-27.at.11.35.27.mp4
ajaxharg commentedon Sep 5, 2026
Adding my experience — this matches what I've been seeing, confirmed via my own logs.
Environment: macOS (arm64), opencode 1.18.25, model
opencode-go/qwen3.8-flash& GLM through ollamaLog evidence (
~/.local/share/opencode/log/opencode.log): 123 occurrences ofAI_APICallError: Internal server errorover Aug 30 – Sep 5, all fromproviderID=opencode-go modelID=qwen3.8-flash, interleaved withError from provider (Console Go): Upstream request failed: Endpoint is unavailable.No other provider in my config shows this error pattern. Errors appear in clusters every few hours — consistent with intermittent upstream backend failures for specific models, as described by earlier commenters.Impact on sessions: When the stream dies mid-run, the errored assistant message is persisted in the session DB, and subsequent turns appear to re-send that broken tail, so the session stays broken even after switching models or restarting the app. Only
/undo(or a new session) recovers it. IMO a provider 500 shouldn't be able to leave a session in a state that survives model switching — worth a defensive fix client-side.One more data point: I've been using opencode daily for quite a few months and never saw this issue until recently — it started appearing in the last few weeks and is now recurring on a regular basis.
thdxr commentedon Oct 11, 2026
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (
opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊