Z.ai has made the GLM-5.3 API available with unchanged pricing, offering stronger coding and long-horizon agent performance compared to its predecessor. Concurrently, Cerebras introduced its new CS-4 computer, which it claims is multiple times faster than its previous generation and more responsive than Nvidia-based systems.
- GLM-5.3 API pricing remains at 1.4/4.4 per million tokens while providing improved model capabilities.
- Cerebras CS-4 is currently being sampled by a small group of customers ahead of wider availability in the third quarter.
- OpenAI temporarily slowed frontier model scaling and paused some reinforcement-learning training due to cybersecurity concerns.
These developments highlight ongoing advancements in AI model accessibility and hardware infrastructure, alongside emerging security challenges in model training.