Data Brief

Evo 2 is the largest, fully open language model in biology

Evo 2 was trained with 40 billion parameters on 9.3 trillion tokens, and its training code, data, and model parameters are fully open-source.

Training FLOPS were estimated for other large models, including Pythia Large and ESM 3, and may not be completely accurate.
Chart: Niko McCartySource: Brixi G. et al.Get the dataDownload image