"DiffusionGemma's speedup is designed for local and low-concurrency inference. In high-QPS cloud serving, autoregressive models can be deployed to saturate compute efficiently, so DiffusionGemma's parallel decoding offers diminishing returns and can result in higher serving costs"

So the diffusion process takes more GFLOPs, if you have enough users you can already balance memory and compute.

Next year, and the year after, Fable, GPT 5.5 and Gemini 3.5 will feel quite ordinary. And perhaps even within reach of a prosumer running models locally.

The thing is, diffusion models perform somewhat worse than autoregressive on text. So you lose some performance.

Speed is the big advantage. Autoregressive when doing local inference is mostly memory bound; you're doing one token at a time, for each token you need to load all weights. MTP helps a bit by allowing you to draft tokens in a smaller model and then verify them in parallel with the larger model, allowing you to do a few computations for every memory load, but because you're still doing tokens sequentially and need to discard invalid drafted tokens, you can only get so much speedup.

For hosted models, however, you can batch many token generations together, fully utilizing all of the compute while no longer being bottlenecked on memory bandwidth. So they are already operating at close to max efficiency.

So, diffusion kind of loses its beneifit in hosted models. Sure, maybe you could pay more to have slightly lower latency responses by doing diffusion for one user at a time instead of autoregressive for many in parallel. But given that it also reduces accuracy, it's hard to see where you'd really want that. Unless they're able to bring it up to par with autoregressive, it seems like it's a bit of a dead out outside of local models where you're generally just doing one thing at a time.

I've had pretty good success with LLMs after putting in place metrics to measure true complexity (not cyclomatic).

Does not need opus level to write, and easy to iterate on.

The mechanism isn't the same as speculative decoding. Speculative decoding happens sequentially and (usually) a couple of tokens at a time; diffusion doesn't, and does blocks of text at once. I haven't read the collateral yet but my assumption would be that it's trained to keep the specific experts stable across a diffusion block.