Open Models Almanac
Methodsmodel architecture

specific method · filed under model architecture

Diffusion Transformer (DiT)

A transformer used as the diffusion model, with no more specific architectural mechanism established by the evidence.

The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See model architecture for the methods filed beside it.

Also called Diffusion transformer architecture, diffusion transformer, Diffusion Transformer action head, Diffusion transformer action denoising.

sources
4
models
2
labs adopt it
2
strongest
core

Documented in

Black Forest Labs

Evidence

4 quotes from 4 sources, one card per model, each opening with its strongest.

How sources treat it
core 4

GR00T N1.7

NVIDIAadopted3 sources

A 32-layer DiT takes the VLM's output and live robot state, then denoises them into precise motor commands in real time.

coremodel architecturein Isaac GR00T N1.7NVIDIA
2 more from this model

diffusion transformer head that denoises continuous actions

coremodel architecturein Isaac GR00T N1.7NVIDIA

implemented as a diffusion transformer (DiT)

coreunclearin Isaac GR00T N1.7NVIDIA
↑