Revolut is using GLM 5.3 to cut coding costs from Claude and Codex. This model is produced by
z.ai. 2000 engineers are served by 56 Nvidia B200 GPUs, with blended cost of around 30c per million tokens.
GLM is short for General Language Model (which are different from the GLMs in statistics, which are Generalized Linear Models, a topic for another post!).
No comments:
Post a Comment