AMD introduces open-source OLMo LLM, aiming to rival AI industry leaders

Outstanding performance and successful benchmark results

Based on internal tests, AMD’s OLMo models demonstrated impressive performance compared to similar open-source models like TinyLlama-1.1B and OpenELM-1_1B in various multitasking and general reasoning evaluations. The company reported a notable performance boost exceeding 15% in tasks related to GSM8k, thanks to AMD’s advanced multi-phase supervised fine-tuning and Direct Preference Optimization (DPO) techniques.

According to AMD, OLMo displayed a competitive advantage of 3.41% in AlpacaEval 2 Win Rate and a 0.97% improvement in MT-Bench during multi-turn chat assessments when pitted against its nearest open-source counterparts.

Nevertheless, in the broader landscape of Language Model (LLM) technologies, Nvidia’s GH200 Grace Hopper Superchip and H100 GPU maintain their dominance, especially in handling extensive, diverse AI workloads. Nvidia’s emphasis on cutting-edge features like C2C link, designed to enhance data exchange efficiency between its CPU and GPU, positions Nvidia as a frontrunner by offering a speed boost for demanding inference tasks such as recommendation systems.

About Author

AndyC

Andy Curtis is an award-winning security consultant, researcher and public speaker. He has been working in the computer security industry since the early 1990s, having been employed by state and federal government, leading healthcare and banking providers across three continents. He has given talks about computer security for some of the world’s largest companies, worked with law enforcement agencies on investigations into hacking groups, and is a regular voice on TV and radio explaining IT security threats.

See author's posts