DBRX

Across a range of standard benchmarks, DBRX sets a new state-of-the-art for established open LLMs. Moreover, it provides the open community and enterprises building their own LLMs with capabilities that were previously limited to closed model APIs.

DBRX is a cutting-edge open large language model (LLM) developed by Databricks, designed to set new benchmarks in performance and efficiency. It surpasses established open models like GPT-3.5 and competes with Gemini 1.0 Pro, excelling in programming, mathematics, and general-purpose tasks. Built on a fine-grained mixture-of-experts (MoE) architecture, DBRX achieves faster inference speeds and higher quality outputs while using fewer active parameters compared to other models. It is pre-trained on 12 trillion tokens of curated data, offering superior token efficiency and training performance. Available on Hugging Face, DBRX empowers enterprises and the open community to build and customize their own LLMs, making advanced AI capabilities accessible to all.