Heroku AI now supports DeepSeek V3.2, GLM 4.7, Kimi K2.5, and MiniMax M2.1 models

Change effective on 18 February 2026

Heroku’s Managed Inference and Agents add-on now supports these AI models:

  • DeepSeek V3.2: An open-weight LLM that supports conversational chat, tool-calling, and high-efficiency reasoning.
  • GLM 4.7: An open-weight LLM that supports conversational chat, tool-calling, and stable multi-step reasoning.
  • GLM 4.7 Flash: An open-weight LLM that supports conversational chat, tool-calling, and low-latency agentic tasks.
  • Kimi K2.5: An open-weight LLM that supports conversational chat, tool-calling, and multimodal agentic workflows.
  • MiniMax M2.1: An open-weight LLM that supports conversational chat, tool-calling, and long-horizon reasoning.

These models are available in the US region and support the /v1/chat/completions API endpoint. Select a model to view its documentation and get started.