Bamba-9B-v2, developed by IBM in collaboration with Princeton, CMU, and UIUC, is an upgraded pretrained model that significantly enhances performance over its predecessor, Bamba v1, by training on an additional 1T tokens. It demonstrates superior leaderboard scores compared to other state-of-the-art models while maintaining a faster inference speed due to its Mamba2 architecture.
bamba ✓
+ ibm
machine-learning ✓
ai-model ✓
performance ✓