jamesob
Featured
local-llm
A comprehensive guide for building and configuring a high-end local machine to run state-of-the-art LLMs, with detailed hardware choices, BIOS tuning, and Docker-based model serving.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It overturns the prevailing LLM-centric VLA paradigm by showing that direct bidirectional vision-language interaction can match or outperform much larger models while drastically reducing computation and memory, enabling real-time deployment on consumer hardware.
Who it is for
- Robotic manipulation researchers
- Embodied AI developers
- Students in robotics and machine learning
- Engineers seeking low-latency policy deployment on consumer GPUs
Use cases
- Real-time robot arm control in simulation benchmarks like LIBERO and RoboTwin
- Multi-task manipulation with a single compact policy
- Low-cost research and prototyping of VLA models on consumer hardware
- Exploring efficient architectures for vision-language-action systems
Strengths
- Extremely fast inference (31.2 ms per step) at 32 Hz on an RTX 4090
- Very low GPU memory footprint (0.9 GB) during inference
- State-of-the-art success rate (97.7%) on LIBERO with only 0.2B parameters
- Open-source code with detailed training and evaluation scripts for two benchmarks
Considerations
- Requires external pre-trained models (DINOv3, BERT, GroundingDINO) and manual data preprocessing (no-op removal)
- Setup involves separate Python environments for different benchmarks (LIBERO vs RoboTwin)
- Evaluated only on simulated manipulation tasks; no real-world robot experiments are reported
README quick start
TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM
Hengyi Xie1*, Chenfei Yao1*, Xianjin Wu1, Xuanyang Xi2, Yiping Tang2, Di Xu2, Yingying Zhu1,
Dingkang Liang1†, Xiang Bai1, Han Ding1
1 Huazhong University of Science and Technology, China 2 Huawei Technologies Co. Ltd, China
- Equal contribution, listed alphabetically by surname. † Project Lead.
This repository contains the official implementation of TurboVLA for the paper TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM.
📄 Abstract
Vision-language-action (VLA) models commonly adopt an LLM-centric V → L → 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 → L → A pathway as a direct V + L → 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.
🔍 Overview
📈 Performance
⚙️ Installation
Clone the repository:
git clone https://github.com/H-EmbodVis/TurboVLA.git
cd TurboVLA
LIBERO and RoboTwin use different simulator and data stacks. We recommend separat
Description
TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM
Related repositories
Similar projects matched by category, topics, and programming language.
slvDev
Featured
esp32-ai
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
gavamedia
Featured
deltafin
Deltafin is a research project that runs the 2.8-trillion-parameter Mixture-of-Experts model Kimi K3 on a single Apple Silicon Mac (e.g., M1 Max with 64 GB) at about 16 seconds per token, using exact, reproducible inference with local or streaming expert loading.