1.6K views Β· 26 likes Β· 3 reposts (at collection)
vLLM Reports Kimi K3 Decode Performance on NVIDIA GB300 NVL72
- Who
- vLLM
- What
- reports DSpark decode performance for Kimi K3 on NVIDIA GB300 NVL72
What happened
Kimi K3 decode on NVIDIA GB300 NVL72: 842 tokens/s at batch 1 with DSpark π A unified CuTe DSL decode path fuses attention, MoE and communication, delivering 1.5β2Γ higher decode throughput in several tested configurations and up to 1.8Γ higher P90 interactivity on the
The story so far
vLLM reported that Kimi K3 reached 842 tokens per second at batch size 1 using DSpark on NVIDIA GB300 NVL72. vLLM said a unified CuTe DSL decode path fuses attention, mixture-of-experts (MoE) operations and communication, delivering 1.5β2Γ higher decode throughput in several tested configurations.
Keep reading
Latest daily Β· Issue 25DeepSeek releases DeepSeek-V4.1-Flash-1017 on its website, app, and APITrending now
- 1DeepSeek Releases V4.1-Flash-1017, Now Available on Ollama Cloud With US Hosting
- 2Meta Removes Business Terms Language That Raised Website Scraping and AI Training Concerns
- 3Google DeepMind Says Four Gemini 4 Argon Agents Beat Minecraft Without Human Help
- 4Anthropic Says Claude Reconstructed Five NIST Elliptic-Curve Seeds Using a Timestamp-Initialized Generator
Tomorrow at 08:00, a briefing like this arrives in your inbox.
The same kind of briefing as the one above: a few stories worth reading. Free. Mail starts after you confirm, and every letter has a one-click unsubscribe. Read a past issue Β· 2026-10-18