Models · MarkTechPost ·
Zyphra Release Zamba2-VL: Hybrid Mamba2–Transformer Vision-Language Models That Cut Time-to-First-Token by About an Order of Magnitude
Zyphra released Zamba2-VL, an open family of vision-language models with 1.2B, 2.7B, and 7B parameters. The models use a hybrid Mamba2 state-space and Transformer backbone, are available under Apache 2.0, and reportedly reduce time-to-first-token by about an order of magnitude versus comparable Transformer VLMs.