1 pointby rene_d3 hours ago2 comments
  • Sphax3 hours ago
    I like that it has 512K context. If it's as good as Qwen 3.8 27b when quantized I could use it just for that alone.
  • mynti3 hours ago
    It scores almost as high as GLM 5.3 and Kimi 3 on Deep SWE, how is that possible?
    • rene_dan hour ago
      The actificial analysis post shows that it does not perform well at all on its newer benchmarks. So probably very benchmark focussed, but still, worth a try for local coding.