# LiquidAI releases block‑diffusion version of LFM2.5 model

> The new 350M‑parameter model denoises 32‑token blocks for faster text generation, matching the original on most benchmarks.

Oossa · 2026-10-03 · https://oossa.com/en/liquidai-releases-block-diffusion-version-of-lfm2-5-model

LiquidAI added a new model called LFM2.5‑350M‑Diffusion‑Exp to Hugging Face on 2026‑10‑03. It works like the original LFM2.5‑350M model but generates text in 32‑token blocks instead of one token at a time. This block‑diffusion approach speeds up decoding, especially at low batch sizes, while keeping accuracy close to the autoregressive version when using eight denoising steps per block. The model uses the same backbone, tokenizer, and chat template as LFM2.5.

## The facts

- Model size: 350 million parameters
- Recommended decode config: 8 denoising steps per 32‑token block

## Why it matters

Developers can get faster responses from a familiar 350 M model without a large loss in quality.

## Sources & references

1. [New model on Hugging Face: LiquidAI/LFM2.5-350M-Diffusion-Exp (text-generation)](https://huggingface.co/LiquidAI/LFM2.5-350M-Diffusion-Exp) – Liquid AI, 2026-10-02

Last updated: 2026-10-03