Description
Environment
- OpenCode version: 1.17.18
- OS: Windows 11
Affected providers
opencode (Zen)opencode-go
Affected models
deepseek-v4-flashdeepseek-v4-promimo-v2.5-pro
Error
When using these models in agent mode, OpenCode shows "Internal server error" and retries repeatedly.
Direct API test shows the backend returns:
{"error":{"message":"Error from provider (Console Go): Upstream request failed","type":"invalid_request_error","code":"invalid_request_error"}}
Verification
- Same request without tools parameter → 200 OK
- DeepSeek official API with tools → 200 OK
- Other Zen/Go models with tools → 200 OK
Working models
- Zen: hy3
- Go: minimax-m3, glm-5.2, qwen3.6-plus, qwen3.7-max
Suspected root cause
OpenCode backend does not correctly handle reasoning_content echoing for DeepSeek V4 tool calls (and possibly MiMo), causing 400 errors that the client displays as "Internal server error".
Expected behavior
Either fix backend handling for these models, or update models.json to mark tool_call: false for affected models until fixed.
### Plugins
opencode-sessions, opencode-context-cache, magic-context
### OpenCode version
1.17.18
### Steps to reproduce
_No response_
### Screenshot and/or share link
_No response_
### Operating System
Windows 11
### Terminal
wezterm，Windows
Description
Environment
Affected providers
opencode(Zen)opencode-goAffected models
deepseek-v4-flashdeepseek-v4-promimo-v2.5-proError
When using these models in agent mode, OpenCode shows "Internal server error" and retries repeatedly.
Direct API test shows the backend returns:
{"error":{"message":"Error from provider (Console Go): Upstream request failed","type":"invalid_request_error","code":"invalid_request_error"}} Verification - Same request without tools parameter → 200 OK - DeepSeek official API with tools → 200 OK - Other Zen/Go models with tools → 200 OK Working models - Zen: hy3 - Go: minimax-m3, glm-5.2, qwen3.6-plus, qwen3.7-max Suspected root cause OpenCode backend does not correctly handle reasoning_content echoing for DeepSeek V4 tool calls (and possibly MiMo), causing 400 errors that the client displays as "Internal server error". Expected behavior Either fix backend handling for these models, or update models.json to mark tool_call: false for affected models until fixed. ### Plugins opencode-sessions, opencode-context-cache, magic-context ### OpenCode version 1.17.18 ### Steps to reproduce _No response_ ### Screenshot and/or share link _No response_ ### Operating System Windows 11 ### Terminal wezterm，Windows