Databricks Model Serving now supports Moonshot AI's Kimi K3 model
Databricks Model Serving now offers Kimi K3, a 2.8-trillion-parameter mixture of experts (MoE) model, as a hosted option. This model includes native vision capabilities and supports a 1-million-token context window. It is designed to enhance long-horizon coding, knowledge work, and advanced reasoning tasks for Databricks users. Developers can access Kimi K3 through the Foundation Model APIs.
Features (1) ›
- Kimi K3 AI model available as a Databricks-hosted model
Databricks Model Serving now supports Kimi K3, a 2.8-trillion-parameter mixture of experts model. This model features native vision capabilities, a 1-million-token context window, and is optimized for complex coding, knowledge work, and reasoning tasks. It can be accessed via the Foundation Model APIs.
https://docs.databricks.com/aws/en/release-notes/product/2026/august#moonshot-ai-kimi-k3-now-available-as-a-databricks-hosted-model
Related releases
- Databricks Unity AI Gateway Now Generally Available Databricks Release Notes ·
- Databricks introduces ai_search() SQL function for RAG pipelines (Beta) Databricks Release Notes ·
- Databricks Delta Lake Sharing with Iceberg Reads Reaches GA Databricks Release Notes ·
- Genie One now available in Databricks Excel Add-in for natural language queries Databricks Release Notes ·
- Databricks Adds Search and Replace for Files in Git Folders Databricks Release Notes ·
- Databricks-managed MCP Connectors Integrate with Unity AI Gateway (Beta) Databricks Release Notes ·