Modal Labs announced Auto Endpoints for GLM-5.3-Flash, enabling developers to deploy and query the 320B-A18B MoE model via OpenAI-compatible endpoints with automated serverless cold-start management.

Key Takeaways

  • ✓Day-0 open-model availability establishes the baseline competitive standard for serverless GPU platforms;
  • ✓Auto Endpoints abstract away containerization and GPU orchestration into immediate OpenAI-compatible calls;
  • ✓Elastic on-demand scaling eliminates idle infrastructure costs for bursty multi-turn coding agent workloads.