Zamba2-VL用混合架构实现高效视觉语言理解，小模型部署更快。
Zamba2-VL Technical Report
- 融合Mamba2与少量Transformer块，构建高效混合视觉语言模型
- 1.2B/2.7B小模型推理速度比主流Transformer快约10倍
- 适合边缘设备部署，性能媲美大模型且效率优势显著
我们提出Zamba2-VL，基于Zamba2——一种结合Mamba2状态空间层与少量共享Transformer块的混合语言模型架构——构建的一系列视觉语言模型。在图像理解、推理、OCR、定位和计数等广泛基准测试中，Zamba2-VL在参数量相近的情况下，性能可与主流Transformer类开源视觉语言模型（如Molmo2、Qwen3-VL、InternVL3.5）相媲美，并显著优于先前基于状态空间模型（SSM）或混合架构的视觉语言模型（如VL-Mamba、Cobra、mmMamba）。继承其骨干网络Zamba2的近线性预填充计算与小而恒定的循环状态特性，Zamba2-VL在相同参数规模下，时间到首个词元（TTFT）比这些基于Transformer的基线模型低约一个数量级，尤其在1.2B和2.7B这类更适用于本地与边缘部署的小规模模型上效率优势更为突出。我们发布了三个模型（1.2B、2.7B、7B）及推理代码，地址为https://huggingface.co/collections/Zyphra/zamba2-vl。
原文摘要 · Abstract (English)
We present Zamba2-VL, a suite of vision-language models built on Zamba2, a hybrid language-model architecture combining Mamba2 state-space layers with a small number of shared transformer blocks. Across a broad range of image understanding, reasoning, OCR, grounding, and counting benchmarks, Zamba2-VL is competitive with leading Transformer-based open-weight VLMs of comparable scale, including the Molmo2, Qwen3-VL, and InternVL3.5 families, and substantially outperforms prior SSM-based and hybrid VLMs such as VL-Mamba, Cobra, and mmMamba. Inheriting the near-linear prefill compute and small, near-constant recurrent state of its Zamba2 backbone, Zamba2-VL delivers roughly an order of magnitude lower time-to-first-token (TTFT) than these Transformer baselines at matched parameter scale, with the efficiency gap most pronounced at the smaller 1.2B and 2.7B scales most relevant to on-device and edge deployment. We release three models -- 1.2B, 2.7B, and 7B -- together with inference code at https://huggingface.co/collections/Zyphra/zamba2-vl.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。