Local Qwen Surprisingly Good for Tasks
DanGrover · x · 2026-07-13
The author switched some tasks usually given to Claude to a local Qwen3.6-35B-A3B, running it via ohmypi.
To their surprise, they found the combination to be "actually good and reasonably fast," noting that trying similar-sized models a few months ago often led to hitting walls and hallucinating tools. This time, the results were noticeably more stable.
More from coding & agent
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21
- A set of agent skills for CAD, robotics, and hardware design — earthtojake · 2026-07-21