Databricks announced the launch of DBRX, a general purpose large language model (LLM) they say outperforms established open source models on standard benchmarks. DBRX democratizes the training and tuning of custom LLMs for enterprises without relying on a small handful of closed models. Paired with Databricks Mosaic AI’s unified tooling, DBRX helps customers rapidly build and deploy generative AI applications that are safe, accurate, and governed without giving up control of their data and intellectual property.

DBRX was developed by Mosaic AI and trained on NVIDIA DGX Cloud. Databricks optimized DBRX for efficiency with a mixture-of-experts (MoE) architecture, built on the MegaBlocks open source project. For a look at model evaluations and performance benchmarks compared to existing open source LLMs like Llama 2 70B and Mixtral-8x7B, and to see how DBRX is competitive with GPT-4 quality for internal use cases such as SQL, visit the Mosaic Research blog.

DBRX is available on GitHub and Hugging Face. On the Databricks Platform, enterprises can leverage its long context abilities in retrieval augmented generation (RAG) systems, and build custom DBRX models on their data. DBRX is also available on AWS and Google Cloud, as well as directly on Microsoft Azure through Azure Databricks.