open-gigaai/giga-world-1

Health v2 · maintenance only

Flags: no_releases young

Adoption not part of the score

1152 stars · 92 forks observed · 2026-09-24

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

GigaWorld-1 is an open-source framework providing training, inference, data processing, checkpoint conversion, and LoRA merge workflows for building robot world models used in robot policy evaluation. It is a roadmap-style project with models and datasets published on HuggingFace.

Use cases

• train a world model for robot policy evaluation
• run inference with a robot world model
• convert and merge model checkpoints with LoRA
• prepare robot demonstration data for world model training
• evaluate robot policies in a learned simulator
• distill a video world model for faster inference

When to choose

• you need to train or fine-tune world models for robotics on Linux GPUs
• you want an end-to-end pipeline from data preparation to inference for robot policy evaluation
• you want to reproduce or extend a published robot world model with open checkpoints and datasets

When to avoid

• you need a general-purpose video generation tool unrelated to robotics
• you lack GPU hardware or a Linux environment
• you need a mature production system rather than a research roadmap with staged releases

Facets

framework · maturity active

machine-learning deep-learning llm-training simulation video-processing gpu-computing robotics machine-learning deep-learning artificial-intelligence simulation python world-models robot-policy-evaluation video-generation lora diffusion-models model-distillation pytorch linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="open-gigaai/giga-world-1")