Kimi K3 Enabled on AMD MI350 and NVIDIA B200 with Day-0 Inference Optimizations
zhyncs42 · x · 2026-07-28
The TokenSpeed team collaborated closely with Moonshot to achieve Day 0 support for the Kimi K3 model on NVIDIA Blackwell (B200/B300) and AMD Instinct (MI350X/MI355X) GPUs in just one week.
Core inference optimization techniques include:
- Prefix caching
- Speculative decoding
- Disaggregated serving
- CUDA Graph decode
The team also shared underlying optimization details for KDA, Gated MLA, Stable LatentMoE, AttnRes, and their unified Flat KV architecture.
Related event: TokenSpeed Enables Kimi K3 Support on NVIDIA and AMD Platforms(2 posts)→
More from Infra
- 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
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11