GLM-5.3 Hits 310 tok/s, Coding Performance Competes with Opus

Yuchenj_UW · x · 2026-09-02

Databricks inference ranks #1 in speed and latency again. Tests show GLM-5.3 reaches an inference speed of 310 tok/s.

On Databricks' internal coding benchmark, GLM-5.3 performs strongly and is considered the strongest open-source coding model currently available, competitive with top-tier models like Fable 5 and Opus 4.8.

Original post →

More from Models

Models channel →