3 ms·
If this runs at Opus 4.5 level for agentic coding then I don't really need any cloud models anymore.
by storus 6mo ago
If this runs at Opus 4.5 level for agentic coding then I don't really need any cloud models anymore.
- senko 6mo agoI tested it with my standard test - HTML/JS minesweeper. From the user POV, the result is great and subjectively equivalent to Opus 4.5. I also asked Claude Code (Opus 4.7) and Codex (GPT-5.4) to review both qwen's output and that of opus 4.5, and both agents concluded qwen's was better. Minesweeper is simple but nontrivial - 600-800 lines of code that need to be internally consistent. At that complexity level, this model is definitely a viable alternative. (haven't tested with planning, debugging and more complex problems yet)
- qudat 6mo agoI just want sonnet 4.5 but I also want it to be fast which is trickier