Z.ai releases GLM-5.3, matching frontier coding benchmarks

Z.ai's GLM-5.3, a ~750B-parameter model, surpasses Moonshot AI's Kimi K3 on many benchmarks and matches Claude Fable 5 or GPT-5.6-Sol on some. It's the same base as GLM-5.2 with extended post-training, available now in the coding plan, with API and open weights on Hugging Face coming in two weeks.