Evo 2 is trained on 9.3T tokens of DNA with single-base resolution at 1M token context length 🥳
We release two models with 7B and 40B parameters along with weights, training and inference code, and pretraining data—making this one of the largest fully open AI models available