Cosmos-Predict2.5

A generative world foundation model that predicts future video frames from text, images or video, aimed at physical AI and robotics.

Licenses

Code: Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit a2c298b (opens in a new tab), on 5 Nov 2026.
Weights: NVIDIA Open Model LicenseTerms of their own, not a standard license
Read on huggingface.co/nvidia/Cosmos-Predict2.5-2B (opens in a new tab).
Commercial use: Unknown

A license Robotheca cannot classify by itself: read it before any commercial use.

Custom weights termsThe weights come under their maker's own terms (NVIDIA Open Model License), not a standard license: read their conditions before any use.

Unifies text-to-world, image-to-world and video-to-world generation in one flow-based model.
Uses Cosmos-Reason1, a vision language model, as its text encoder.
Offers 2B and 14B base checkpoints, pre-trained and post-trained, plus a distilled 2B text-to-world variant.
Includes specialized checkpoints for autonomous driving with seven camera views and for robots, including action-conditioned and policy models.
Supports post-training, LoRA fine-tuning and distillation, with guides for each.
Provides Diffusers support and inference example notebooks.

Written by Robotheca from the project's README (opens in a new tab) at commit a2c298b.

Checked on GitHub on 5 Nov 2026.

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.