Mercury3 diffusion LLM claims 1,000+ tok/s and 54ms first-token latency
大模型之路 · wechat · 2026-10-22
Inception launched Mercury3 (Oct 19, 2026), a diffusion LLM claiming 1,000+ tokens/sec, 54ms first-token latency, and 1M-token context at $0.10/$0.35 per million tokens (closed API, no weights). The piece explains why diffusion decoding's latency is step-count-bound rather than length-bound — its sweet spot being long outputs and interactive apps — while flagging real trade-offs: denoising steps as a rigid quality/speed knob, an ecosystem needing retooling, closed weights, and vendor-reported benchmarks with no third-party replication. Verdict: treat it as a latency-layer option, not an autoregressive killer.
More from Infra
- Bloom Energy Q3 revenue hits $1.24B, up 138% YoY, raises FY26 guidance to $4.4-4.6B — Beth_Kindig · 2026-10-23
- 1M+ PHP services patched in 24h, but ~40M still exposed to PHP2Shell CVEs — FanaHOVA · 2026-10-23
- Trader dismisses AI circular financing theory: every new megawatt booked at rising spot prices — JOBhakdi · 2026-10-23
- Rumor: New Architectures Could Cut Cerebras K3 Wafer Needs From 60 to 10 — bookwormengr · 2026-10-23
- Nvidia plunges 7.21% to $212.31 in a day, still holding a $5.2T market cap — michalmalewicz · 2026-10-23
- Serverless Is a Billing Model: How to Price Stateful Compute by Usage — s4chinraja · 2026-10-23