TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM
2026-07-29 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionRobotics
AI summaryⓘ
The authors propose TurboVLA, a new way to connect vision, language, and robot actions that skips using a large language model as a middle step. Instead, TurboVLA directly combines visual and language inputs and uses a small decoder to predict robot actions, which makes it faster and less computer-demanding. They tested it on a robotic task set called LIBERO and found it works as well or better than bigger models while using fewer resources. This shows TurboVLA can be a simpler and more efficient alternative to current methods in robotic manipulation.
Vision-language-action modelsLarge language modelsRobotic manipulationBidirectional vision-language interactionInference latencyModel parametersLanguage instructionsVisual observationsCompact decoderLIBERO dataset
Authors
Hengyi Xie, Chenfei Yao, Xianjin Wu, Xuanyang Xi, Yiping Tang, Di Xu, Yingying Zhu, Dingkang Liang, Xiang Bai, Han Ding
Abstract
Vision-language-action (VLA) models commonly adopt an LLM-centric $V \to L \to A$ pathway, where visual observations are projected into the representation space of a large language model before being decoded into robot actions. Although effective, this design incurs substantial computation and memory overhead at every policy invocation. In this work, we introduce TurboVLA, a new VLA paradigm that reformulates the conventional $V \to L \to A$ pathway as a direct $V + L \to A$ mapping. Instead of using a large language model as the central interface between perception and action, TurboVLA independently encodes visual observations and language instructions, directly exchanges information between them through lightweight bidirectional vision-language interaction, and predicts continuous action chunks with a compact decoder. This simple design constructs task-conditioned representations directly from visual and linguistic features, significantly reducing the computational and memory costs of VLA inference. On LIBERO, TurboVLA achieves 97.7% average success with only 0.2B parameters, 31.2 ms inference latency, and 0.9 GB inference VRAM on a consumer-grade RTX 4090, matching or outperforming substantially larger VLA policies. These results establish TurboVLA as a simple and effective alternative to the prevailing LLM-centric VLA paradigm, offering a new perspective on how vision, language, and action can be connected for efficient robotic manipulation. Code is available at https://github.com/H-EmbodVis/TurboVLA.