diffusion-models

Diffusion LLMs Are Fast. Your Agent Is Still Slow.

Nemotron-style diffusion LLMs cut decoding time, but agent latency lives in retrieval and tool calls. Here's what actually changes in production.