Topic
Text generation by iterative denoising or unmasking rather than left-to-right next-token prediction, including masked and block diffusion variants.
SoloQ, a calibration-free quantizer for diffusion LLMs, cuts inference FLOPs about 45%, according to a dev.to post on an arXiv paper dated 2026-10-05. That write-up is the only source, and its cost numbers come from a scenario with an unnamed provider.
Reality
Mercury 2.5's diffusion decoder refines many token positions at once, which is why its headline throughput moves with batch size, output length and the reasoning setting Inception ships as an adjustable knob.
Reality
Inception's diffusion model is the fastest endpoint in the cheap tier on both figures, but OpenRouter's median sits at less than half the vendor number and only about 15 percent above Gemini 3.5 Flash-Lite's measured 382 tok/s.
Reality
DFlash uses a diffusion model to draft three tokens at once and claims to beat EAGLE-3. On one consumer GPU running llama.cpp on a JavaScript coding task, it trailed the drafter Google ships with Gemma-4-12B-it.
Reality
Mercury reports over 1,000 tokens per second per user, and Nemotron Diffusion claims 2 to 8 times autoregressive throughput. Both are vendor figures. The sampler that produces them charges for its speed in arithmetic.
Reality