SGLang adds day-one Kimi K3 support and reports 423 tokens/s on GSM8K
BanghuaZ · x · 2026-07-28
SGLang announced day-0 support for Kimi K3, claiming a speed of 423 tokens/s on GSM8K and saying RL support is ready.
- The post describes K3 as a 2.8T-parameter open model with a 1M context window.
- SGLang says it natively implemented and deeply optimized K3’s architecture with fused KDA decode kernels, DP attention, DSpark, PD disaggregation, and KDA-aware prefix caching.
- It reports production readiness after passing the Kimi Vendor Verifier.
- The stack is said to run across NVIDIA GB300/B300/B200/H200/H20, AMD MI350X/MI355X, and more.
- The launch video was also generated by K3 itself and turned into a playable mini game.
Related event: SGLang Day-0 Support for Kimi K3 Boosts Throughput to 423 tok/s(8 posts)→
More from Infra
- Random Engram Reads Kill SSDs: NVLink-Backed CPU Memory and Redis Eyed for Sparse Memory Lookup — bookwormengr · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11