1 points | by rene_d 3 hours ago ago
3 comments
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.
It scores almost as high as GLM 5.3 and Kimi 3 on Deep SWE, how is that possible?
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.
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.
It scores almost as high as GLM 5.3 and Kimi 3 on Deep SWE, how is that possible?
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.