From my evaluation[1] neither Kimi K3 nor Qwen 3.8 are as good as GLM 5.2 at coding. I wonder if there's a marketing gap that's got people underestimate it. It

From my evaluation[1] neither Kimi K3 nor Qwen 3.8 are as good as GLM 5.2 at coding. I wonder if there's a marketing gap that's got people underestimate it. It uses twice as much tokens to achieve the same but the results are significantly better and because it's so much cheaper it's the most economical choice too. [1] https://blog.bosun.ai/software-maintenance-with-open-weight-models-part-1/ https://blog.bosun.ai/software-maintenance-with-open-weight-...