GEN-1.5

1.5 · Family: GEN

Generalist AI's embodied foundation model (VLA) that learns new physical tasks one-shot from a single demonstration and outputs 100 Hz action trajectories.

✓ ActiveRobotics foundation modelVision-Language-Action modelMultimodal📁 GEN

Context window

~30 s wideo

tokens

Release date

19 August 2026

🏢Generalist AIProducer

Overview

GEN-1.5 is an embodied foundation model (Vision-Language-Action) developed by Generalist AI, introduced on August 19, 2026. It processes video input (a context window of roughly 30 seconds) together with sensor data, proprioceptive (robot-state) data and language, and outputs action trajectories at 100 Hz.

According to the company, the model shows the beginnings of one-shot physical-skill learning: it can acquire a new task from a single 3–12 second demonstration, and adapt few-shot via 1–10 gradient steps on 1–5 minutes of data. Reported results are an average 59% (±10%) success with one-shot in-context prompting and 83% (±9%) after few-shot fine-tuning (10 gradient steps), with weight changes below 0.15%.

The model was trained via scaled pretraining on large physical-interaction datasets collected in homes, warehouses and factories, with a stated continuous pretraining run of 8+ months. GEN-1.5 belongs to the GEN model family (successor to GEN-1).

Classification

Robotics foundation modelVision-Language-Action modelMultimodal

Family: GEN

Applications

Robotic manipulation

Access & deployment

Weights: Closed

Key parameters

📏 Context: ~30 s wideo

📥 Input: video, robot sensors, robot state data, text

Robotics

Dexterous manipulationRobot manipulationRobot controlEmbodied task planningMotion planningObject affordance understanding

Technical specification

Context window

~30 s wideo

tokens

Modalities

⬇ Input

videorobot_sensorsrobot_state_datatext

⬆ 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

Video understanding

The model's ability to analyse and interpret video content — recognising actions, motion, events and relationships between objects over time.

Category: video

Zero-shot learning

The model's ability to perform a new task without dataset-specific training or hyperparameter tuning — prediction is produced in a single pass from context.

Category: other

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

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

Planning

Forming and executing action plans for complex tasks.

Category: planning

Robotics

Dexterous manipulationRobot manipulationRobot controlEmbodied task planningMotion planningObject affordance understanding

Application domains

Robotic manipulation