Skip to content
← Speculative & Constrained Decoding / 21
expertNewGoogle DeepMindBasetenSnowflake

A diffusion LLM emits 15 to 20 tokens per forward. Is it 15x faster than autoregressive decode? Derive where it wins.

Tokens per forward pass is the number on the model card. The answer prices what each pass costs, finds why a mixture-of-experts canvas reads every expert, and locates the batch where autoregressive decode takes the GPU back.

Updated Oct 2026 · Worked answers on inference mechanisms, with calculations, implementation details, and follow-up questions.

Tokens per forward pass is the number on the model card. The answer prices what each pass costs, finds why a mixture-of-experts canvas reads every expert, and locates the batch where autoregressive decode takes the GPU back.

10 answers per topic instead of 5, plus saved progress and bookmarks · costs nothingor unlock all 213 remaining answers · $25

Every worked answer and derivation in all 12 topics, every concept page, the full practice-test bank, and every code drill’s edge-case tests and reference solution.

6 months · One payment · No auto-renewal

UP NEXT ON YOUR JOURNEY
Next in this trackAn RL rollout engine speculates with an EAGLE head trained on the step-0 policy, and rollouts are no faster now. Why?Popular right nowWhy is batch-1 decode on a 70B model bandwidth-bound, and what is its latency floor on two H100s?Popular right nowDecode attention intensity is fixed at the query-to-KV head ratio. Derive that, then tell me what MLA does to it.
DISCUSSION · 0

Nobody has commented on this answer. Add a correction or another way to work it.