Model
Diffusion Gemma
A text generation model from Google based on diffusion models that processes multiple tokens in parallel at once, reducing latency and making it suitable for low-latency scenarios.
Timeline (last 90 days)
Oct 30
Diffusion models (such as Diffusion Gemma) process blocks of 256 tokens in parallel at a time, whereas autoregressive models predict tokens one by one.
Unverified50%