LLaDA-1.5 is a 1.5-scale diffusion language model that generates text through a denoising diffusion process rather than autoregressive next-token prediction. It is intended for research into alternative generative modeling paradigms for language.
LLaDA sits in PulseGate's Foundation models & chat category. It focuses on exploring non-autoregressive text generation using diffusion techniques as an alternative to traditional transformer decoders. It is built as an open-source project for developers. The project is open source (Open Source). LLaDA is available on the web, the command line, and API.
Behind LLaDA is GSAI-ML, and it first shipped in 2025. The project is developed in the open on GitHub with 55 stars.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do