OpenCode silently sends wrong top-p for Qwen models, hurting output quality
JadedSession · reddit · 2026-08-18
A Reddit user found that coding agent OpenCode always sends top-p=1.0, while Qwen models expect 0.95 (thinking) or 0.80 (no-thinking). The bug is invisible in llama.cpp and derivatives, but ninfer logs sampler settings per request and auto-configures correct ones.
The commit that introduced the change was nominally for another fix with no explanation. There's no per-model/provider override; the workaround is setting "build": {"topp": 0.95} in the agent config, which affects all models.
More from coding & agent
- Meta open-sources Muse Spark Open agent model with day-0 TPU support in SGLang — BanghuaZ · 2026-10-16
- Saying "load-bearing" in your prompt makes Opus 5.5 stop holding back — zacharynado · 2026-10-16
- You.com partners with Paperclip to give every agent live web access — PolarBearby · 2026-10-16
- I priced every Grok Bot call: $20 to $200 plan in 5 days, long jobs eat 52% — jonathanmalkin · 2026-10-16
- GitHub is criminally underrated as a free data store, devs argue — blelbach · 2026-10-16
- Claude Desktop doesn't support MCP sampling — here's a workaround for two-way MCP apps — wuj · 2026-10-16