LingBot-VLA

4B / 6B

A 4B/6B vision-language-action foundation model for robot control, pre-trained on 20,000 hours of real-world data from 9 dual-arm robot configurations.

✓ Active✓ Public access⚖ Open sourceVision-Language-Action modelRobotics foundation model

Parameters

4B / 6B

parameters

Release date

1 January 2026

Access:DownloadDeployment:💻 Local

Overview

LingBot-VLA is a Vision-Language-Action (VLA) foundation model built for robotic control. It comes in 4B and 6B parameter variants (plus a LingBot-VLA-4B-Depth version with depth sensing) and is built on top of the VeOmni project.

The model stands out for its pragmatic data approach: pre-training used 20,000 hours of real-world data collected from 9 popular dual-arm robot configurations. It integrates vision, language, and action modalities, mapping observations and instructions to robot actions.

The developers report clear superiority over competing models on both simulation and real-world benchmarks, along with a 1.5–2.8x training speedup over existing VLA codebases (depending on the base vision-language model). The weights are released under the Apache 2.0 license.

Classification

Vision-Language-Action modelRobotics foundation model

Access & deployment

Download

Local

Weights: Open source

Key parameters

🧩 Parameters: 4B / 6B

✓ Fine-tuning

📥 Input: image, text, depth, robot state data

Technical specification

Parameters

4B / 6B

parameters

License

Apache 2.0

Hardware requirements

4B and 6B variants (safetensors). Built on the VeOmni framework; training is 1.5-2.8x faster than typical VLA codebases.

Features:✓ Fine-tuning

Modalities

⬇ Input

imagetextdepthrobot_state_data

⬆ Output

robot_actionsmotion_trajectories

Capabilities and applications

Native model capabilities

Vision-language-action grounding

The ability of a VLA model to ground visual perception and a language instruction into a concrete physical robot action. The model understands the scene and intent, then generates an executable action sequence, closing the loop from observation to motion.

Category: robotics

Action conditioning

Controlling model generation via action signals (camera, robot pose, commands, speech) rather than text prompts alone.

Category: multimodal

Cross-embodiment transfer

The ability of a single model to control robots with different morphologies (humanoids, dual-arm rigs, mobile platforms) without training a separate model per platform. Intelligence is decoupled from embodiment, so the same policy runs on hardware with different kinematics and dynamics.

Category: robotics

Instruction following

Precisely following instructions contained in the prompt: response format, length, style, constraints (e.g. 'reply in six words'). GPT-5.1 significantly improved this capability compared to GPT-5.

Category: language

Real-time inference

The model's ability to generate responses with very low latency (>1000 tokens/sec) on specialized inference hardware (e.g. Cerebras WSE), enabling interactive, turn-by-turn collaboration with a human.

Category: coding