As of right now I think GLM 5.3 Flash is one of the best models to run locally. For me it beats Qwen 3.8 Flash Next on every coding task I throw at it.
Not only does it beat it, it also doesn't loop or get stuck in thinking. While Qwen 3.8 Flash Next may be faster it has taken anywhere from 2-4x longer to get the same tasks done.
This is why Qwen 4 is so important right now because it would be a step up.