SGLang Summit agenda: LinkedIn serves ads model at 10K+ QPS under 80ms
DavidBennett__ · x · 2026-11-13
SGLang Summit 2026 runs Nov 12-13 at Fort Mason, San Francisco, and the full two-day agenda is out. Highlights:
- Production: Cloudflare on large-scale deployment; LinkedIn explains serving a small ads-ranking model at 10K+ QPS with scoring latency under 80ms.
- Engine: Meta bringing PyTorch to SGLang; Alibaba extending HiCache for long-running agents.
- Agentic engineering: Perplexity's minimalist trainer for trillion-parameter models built with coding agents; Modal's early results using agents to develop and manage SGLang inference services.
- Training: Thinking Machines Lab and Meta on TorchTitan; Decagon using RL to improve instruction following and cut hallucinations in customer-facing agents.
- Hardware: NVIDIA on porting SGLang to Vera Rubin; Google on SGLang-TorchTPU for global-scale production inference.
- Real-time & media: SGLang Omni for interruptible conversations; HeyGen and MiniMax on GPU cost optimization and film workflows.
More from coding & agent
- Dev open-sources first MCP server for Power BI custom visuals, seeks feedback — Constant_Plane4937 · 2026-11-13
- LLMs' mental map of SF tested: best street-level score 0.32 vs 0.23 random — Evening_Hawk_7470 · 2026-11-13
- Levels 9 and 10 of AI adoption: simulating strategies with competing agent organizations — danshipper · 2026-11-13
- Idea basins: why autonomous agents get stuck, and how frontier models explore differently — AlexGDimakis · 2026-11-13
- Cap, the open-source screen recorder, launches on iOS with instant share links — JosephJacks_ · 2026-11-13
- Why reasoning distillation fails on 9B models: six LoRAs, mixed results — SignificantZebra5883 · 2026-11-13