Colendi AIResearch
Menu

Inference & Serving · 2025

Mercury: Ultra-Fast Language Models Based on Diffusion

Inception Labs (Samar Khanna, Siddhant Kharbanda, Shufan Li, et al.)

Primary source

First commercial-grade diffusion LLM, generating tokens in parallel to reach an order-of-magnitude higher throughput than autoregressive coders at comparable quality.

Editorial record

Plain-language summary

First commercial-grade diffusion LLM, generating tokens in parallel to reach an order-of-magnitude higher throughput than autoregressive coders at comparable quality.

Knowledge graph

Relationships

Descendants

  • Depends onEvidence: Strongly supported

    Attention Is All You Need

    Builds on the inference serving lineage in the archive

    freshness sweep 2026

Source record

Provenance

Record ID
P-616
Record created
2026-07-13
Last reviewed
2026-07-14
Record version
2

Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.