DiffusionGemma: 4x faster text generation
Get smart on it
DiffusionGemma is an experimental open model that generates text up to 4 times faster than standard language models by creating entire blocks of text simultaneously instead of one word at a time. Rather than processing text sequentially like a typewriter, it drafts 256 tokens in parallel, which fully utilizes a dedicated GPU's processing power during local inference. The model is designed for speed-critical interactive tasks such as in-line editing and code infilling, though it prioritizes speed over output quality compared to standard production models. This approach works best for local, single-user applications where traditional language models leave hardware underutilized while waiting for the next token prediction.
Related news
Vibe coding platform Base44 launches own model as AI startups seek defensibility
Wix-owned vibe coding platform Base44 has started rolling out its own AI model — with hopes that it will eventually outperform frontier models.
DiScoFormer: One transformer for density and score, across distributions
DiScoFormer is a transformer model that estimates both the density and score of a data distribution in a single forward pass without requiring retraining for new distributions. The density describes where data points cluster, while the score, the gradient of log-density, points toward more probable regions and is used in diffusion-based generative models, Bayesian sampling, and scientific simulations. Existing methods force a trade-off: classical kernel density estimation works on any distribution but loses accuracy in high dimensions, while neural score-matching models stay accurate in high dimensions but must be retrained for each new distribution. DiScoFormer significantly outperforms kernel density estimation, cutting score error by 6.5 times and density error by more than 37 times in 100 dimensions, while generalizing to distributions with shapes and complexities not seen during training.
