vLLM dev: software optimization can yield superlinear gains, up to 80x in one generation

woosuk_k · x · 2026-11-17

Responding to abstract AI supply-vs-demand stock analyses, Woosuk Kwon (vLLM) argues real-world performance is nonlinear to FLOPS, HBM bandwidth, and model size. Within a single hardware generation, architecture and software optimizations matter enormously — e.g., DeepSeek v4.1 Flash can deliver 80x higher performance with the right optimizations, and superlinear gains are achievable.

Related event: vLLM Team Counters Abstract Compute Supply-Demand Estimates With 90x Throughput Gains(2 posts)→

Original post →

More from Infra

Infra channel →