AMD Open-Source 1B OLMo Language Models

AMD introduces its first series of 1 billion parameter language models, AMD OLMo, offering significant AI capabilities and fully open-source for community collaboration. These models are trained on AMD Instinct GPUs using trillions of tokens and demonstrate superior language understanding and reasoning.

• AMD released the AMD OLMo language model series.
• OLMo is trained on Instinct MI250 GPUs.
• The models consist of 1 billion parameters.
• The training used 1.3 trillion tokens.
• AMD open-sourced OLMo's training details.