Meta slide shows SGLang serving billions of Muse and Meta app users across four chip types
BanghuaZ · x · 2026-11-14
SGLang core team member Banghua Zhang shared a Meta talk slide revealing SGLang runs in production serving billions of users across the Muse app, Muse Code, Model API, and Meta family of apps (Instagram, Facebook, WhatsApp, Messenger), with multi-cluster routing on top of NVIDIA GPUs, AMD GPUs, Google TPUs, and Meta's own MTIA accelerators.
More from Infra
- Musk predicts agentic AI traffic will hit 1,000x human traffic in 5 years; agent traffic already exceeds human — rohanpaul_ai · 2026-11-14
- AWS Emulator Goes GA: 200+ Services Run Locally, Built Almost Entirely by AI Agents — iamfakhrealam · 2026-11-14
- FAA, Microsoft 365, Claude, Workday All Broke in 48 Hours — No Hackers Needed — r0ck3t23 · 2026-11-14
- llama.cpp ASR/TTS support roundup: Qwen3-ASR/TTS landing, audio output still in planning — Kahvana · 2026-11-14
- Beff Jezos floats Jensen Huang running TSMC and joining forces with Musk's Terafab — beffjezos · 2026-11-14
- vLLM serves Qwen3.8-Flash FP4 on a Mac at ~40 tok/s for local agents — alfcnz · 2026-11-14