Mainly missing GPT-5.3/5.4 which greatly improved output quality (but not speed or cost) - depending on your use-case it beats Opus 4.6
FWIW, I can attest to GLM-5 being by far the best performing open weights model right now for analysis and review. It has been a good tool, and often complements GPT and Claude in a multi-round analysis/review setting.
Mainly missing GPT-5.3/5.4 which greatly improved output quality (but not speed or cost) - depending on your use-case it beats Opus 4.6
FWIW, I can attest to GLM-5 being by far the best performing open weights model right now for analysis and review. It has been a good tool, and often complements GPT and Claude in a multi-round analysis/review setting.