GLM-5.3 Hits 310 Tok/s on Databricks to Top Open Source Coding Benchmarks

Zai_org’s open-weight GLM-5.3 and GLM-5.3-Flash models integrated into the Databricks, Arcee, and Perplexity platforms this week. Databricks reported an inference speed of 310 tok/s for GLM-5.3, which it ranked as the strongest open-source model for coding based on internal benchmarks.

The Perplexity Agent API expanded its offering to 42 models with the addition of the GLM-5.3 release. Arcee incorporated both the standard and Flash variants into its API catalog, while Databricks noted that the model's capabilities are competitive with Fable 5 and Opus 4.8.