VLA Gallery / All models / New in 2026

G0Plus

Galaxea AI (OpenGalaxea) · 2026-01

VLANew in 20263BNot disclosed

Galaxea's January 2026 public checkpoint (G0Plus_3B-base), pretrained on over 2k hours of real-world robot data.

Paper

No paper · Related Galaxea Open-World Dataset and G0 Dual-System VLA Model

Overview

Category

New in 2026

Type

VLA

Released

2026-01

Organization

Galaxea AI (OpenGalaxea)

Open weights

Yes

Parameters

3B (README)

Architecture

Backbone LLM

Not disclosed

Vision encoder

—

Action head

Not disclosed (Not disclosed)

Chunk

Not disclosed

Control rate

Not disclosed

Architecture (original summary)

Not disclosed

Training

Pretraining accelerators

Not disclosed

Pretraining duration

Not disclosed

Pretraining steps & batch

Not disclosed

GPU/TPU-hours

Not disclosed

Fine-tuning

Not disclosed

Data type

Real robot (real-world robot data)

Data scale

Over 2k hours (2k+ hours)

Main datasets

Not disclosed

Tasks & evaluation

Not disclosed

Robot

Not disclosed

Inference

Inference device & speed

Not disclosed

Inference memory

Not disclosed

Weights bf16 (calc.)

6 GB

Where it runs

Not disclosed

Reinforcement learning & base

Reinforcement learning

Not used · Real-robot data pretraining (supervised)

Base control

Wheeled (Direct base velocity/pose output) · R1 Lite actions directly output chassis velocity (chassis.velocities)·torso velocity together with arms·grippers (official fine-tuning config)

Notes

Key features

Galaxea's January 2026 public checkpoint (G0Plus_3B-base), pretrained on over 2k hours of real-world robot data. Its successor is G0.5.

Notes

Released 2026-01-04 (README). Commits from 2026-01-04 to 2026-06-16 fall under the 'G0 PLUS Community License Agreement' (the three-tier license on the HF G05 card). The note does not describe architecture or benchmarks.

Verification

Confirmed