Z.ai (GLM)
External APIs
Z.ai serves the GLM range. GLM 5.3 Prime costs $2.80 and $8.80 per million, and the editor publishes no retention period for API inputs.
Key points : GLM 5.3 Prime · Open weights available · Retention not published.
Models served : the range opens with GLM 5.3 Prime, released on 23 September 2026, at $2.80 and $8.80 per million for a 1,000,000-token context, and GLM 5.3 FlashX at $0.37 and $1.25 for volume workloads. The weights of GLM 5.3 and its Flash variants are published.
Where the data goes
The privacy policy states that models are trained on publicly available internet sources, without saying what happens to API inputs. The terms of use refer to additional terms for API services that are not published at that address. This finding of absence was verified on both documents on 18 October 2026.
Retention and training
No retention period is published for inputs, and nothing says whether they feed training. The terms do, however, forbid the customer from training a competing model from the service. That asymmetry is worth knowing before entrusting it with anything other than public data.
When to use it from a sovereign platform
The GLM range holds up in quality at an intermediate price. As with Qwen, the published weights offer a way out: as long as input retention is undocumented, sensitive data belongs on-premises, not in the API.