Inkling Model Lands on Databricks

Thinking Machines Lab's Inkling, an open-weights AI model, has arrived on the Databricks platform. This integration allows enterprises to leverage the model for coding and agentic reasoning tasks directly on their own data.

The Inkling model Databricks integration positions the model within Databricks' Unity AI Gateway. This provides a centralized environment for managing security, permissions, cost controls, and observability for AI deployments.

Inkling is designed to excel at coding workflows and supports multimodal inputs, making it versatile for various applications. Its availability as an open-weights model means it can be fine-tuned on proprietary datasets for specialized, high-accuracy tasks.

Enterprise Control and Flexibility

Databricks emphasizes the governance benefits, stating that data remains within the enterprise's controlled environment. This addresses key concerns around data privacy and security when deploying advanced AI models.

The platform offers flexibility, allowing teams to choose and customize models. This avoids vendor lock-in and enables selection of the best model for specific workloads, whether open or proprietary.

Cost optimization is another advantage, as open-weights models bypass per-token API pricing common with proprietary alternatives. This allows for more predictable and scalable inference spend.

Accelerating Development

Developers can connect Inkling to coding agents like Cursor and OpenCode through the Unity AI Gateway. This aims to streamline the development of AI-powered coding assistants and applications.

Databricks provides multiple avenues for users to get started, including an AI Playground for experimentation and direct deployment through the Unity AI Gateway. Integration with Agent Bricks is also supported for building complex, data-driven agents.

The Inkling model is now live on Databricks, with further support for SQL querying anticipated soon.