Databricks has launched the DBRX GenAI large language model (LLM) to allow customers to build, train, and serve their own custom LLMs more affordably. The open-source model, DBRX, outperforms closed models like ChatGPT and GPT-3.5 on industry benchmarks and is faster in terms of tokens per second. It uses a mixture-of-experts architecture and is optimized for efficiency.
DBRX has been tested on various benchmarks, including the Massive Multitask Language Understanding (MMLU) benchmark, Programming (Human Eval) challenge dataset, and Math GSM8K dataset. Results show that DBRX performs well compared to other open-source LLMs and GPT-3.5.
The model is freely available for research and commercial use on GitHub and Hugging Face. Databricks is hosting a webinar to introduce DBRX and its capabilities. Customers like Nasdaq are excited about the release of DBRX due to its strong performance and cost-effective serving economics.
However, relying solely on the speed and performance of a specific GenAI model may not be sustainable in the long run, especially without continuous development. It is important to consider both hardware and software advancements in the field of AI to ensure long-term success.
Source link
Source link: https://blocksandfiles.com/2024/03/27/databricks-claims-lead-in-gen-ai-model-performance/