Zyphra Release Zamba2-VL: Hybrid Mamba2–Transformer Vision-Language Models That Cut Time-to-First-Token by About an Order of Magnitude
Quick Answer
Zyphra has unveiled Zamba2-VL, a series of open vision-language models with 1.2B, 2.7B, and 7B parameters.
Quick Take
These models utilize a hybrid Mamba2 and Transformer architecture, achieving a significant reduction in time-to-first-token by approximately an order of magnitude, while remaining competitive with existing Transformer .
Key Points
- Zamba2-VL models come in 1.2B, 2.7B, and 7B parameter sizes.
- The models are built on a hybrid Mamba2 state-space and Transformer backbone.
- They are released under the Apache 2.0 license.
- Zamba2-VL achieves a time-to-first-token reduction by about an order of magnitude.
- These models remain competitive with similar Transformer vision-language models.
Source Excerpt
Zyphra has released Zamba2-VL, a family of open at 1. 2B, 2. 7B, and 7B parameters. The models use a hybrid Mamba2 state-space and Transformer backbone, shipping under Apache 2. 0. They stay competitive with comparable Transformer VLMs while cutting time-to-first-token by about an order of magnitude. The post Zyphra Release Zamba2-VL: Hybrid Mamba2–Transformer Vision-Language Models That Cut Time-to-First-Token by About an Order of Magnitude appeared first on MarkTechPost.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from MarkTechPost
See more →Meet Flash-KMeans: An IO-Aware, Exact K-Means That Runs Over 200× Faster Than FAISS on GPUs
Flash-KMeans is an open-source, IO-aware k-means implementation that operates over 200× faster than FAISS on NVIDIA H200 GPUs. It achieves 17.9× end-to-end and 33× speedup over cuML by optimizing distance calculations and updating mechanisms without approximating results. This advancement significantly enhances performance for data scientists and machine learning practitioners.