Ai2's Olmo 2 1B Model Outperforms Rivals from Google and Meta

The Allen Institute for AI (Ai2) has released Olmo 2 1B, a 1-billion-parameter AI model that outperforms similarly-sized models from Google, Meta, and Alibaba on several benchmarks, announced on their website. Olmo 2 1B is available under the Apache 2.0 license on Hugging Face, and Ai2 has provided the code and datasets used for its development.

Olmo 2 1B is designed to be accessible, running efficiently on consumer-grade hardware like modern laptops and mobile devices. It was trained on a dataset of 4 trillion tokens from various sources, enabling it to excel in arithmetic reasoning and factual accuracy tests, outperforming Google's Gemma 3 1B, Meta's Llama 3.2 1B, and Alibaba's Qwen 2.5 1.5B.

Despite its capabilities, Ai2 cautions that Olmo 2 1B can produce problematic outputs, including harmful or sensitive content, and recommends against its use in commercial settings.