$IREN / FireworksAI: Kimi K3 Incredible Demand Day 0 FireworksAI has the best performance but severely out of GPUs: 1. Fireworks has 145% Throughput of MoonShot AI at 57% the latency. 2. Fireworks has 228% Throughput of Nebius at 56% the latency. GPU shortage is absolutely hammering FireworksAI uptime as it's uptime dropped for 99.99% at the beginning of the morning to 99.29% mid morning to 84.94% now! This is partially because FireworksAI is the ONLY AI Platform to provide both managed inference and managed training for Kimi K3. Managed training takes alot more GPUs than managed inference. FireworksAI needs to increased their contract capacity with $IREN ASAP! Credits for Live Stats: @JordanNanos
在 X 查看 ↗