8 engineers on Hacker News describe using vLLM first-hand: 3 endorse it, 1 regret it, 0 are mixed.
“vLLM gang all the way! For now at least, as it's what we have most experience with, and we find it the most flexible.”
endorse · production · TofuLover · 2026-08-23 · source ↗ · wrong label?“We run vLLM in a container with hardware passthrough for gitops, having the entire environment in a single container would drastically improve things”
regret · production · hadlock · 2026-08-03 · source ↗ · wrong label?“vLLM is better for Tensor Parallel and also better for batched inference... We run devstral fp8 on 2x A6000”
endorse · production · NitpickLawyer · 2025-07-23 · source ↗ · wrong label?“I've deployed the OpenAI-compatible completions API over Llama2 in production with vLLM, and it works perfectly.”
endorse · production · sbrother · 2023-11-18 · source ↗ · wrong label?“Compared with the alternatives we measured (vLLM with prefix caching, and dedicated decision models)”
neutral · personal · giarec · 2026-10-30 · source ↗ · wrong label?“I worked on the vLLM feature”
neutral · production · mmastrac · 2026-10-26 · source ↗ · wrong label?“we run on B200, the official weights with specific configurations that we have on vllm.”
neutral · production · erans · 2026-10-10 · source ↗ · wrong label?“For this reason we run vLLM on a separately sandboxed VM on a firewalled VLAN.”
neutral · production · angry_octet · 2026-08-24 · source ↗ · wrong label?