CoreWeave Leads Artificial Analysis Kimi K2.6 Benchmark
CoreWeave Blog, Monday, May 11th, 2026
CoreWeave achieves #1 ranking for Kimi K2.6 inference speed and price-performance using NVFP4 and EAGLE3 optimization.
CoreWeave has achieved the top ranking in Artificial Analysis benchmarks for Kimi K2.6 inference speed and price-performance, delivering 205 output tokens per second with the best cost efficiency among 11 providers. The company optimized performance through training custom NVFP4 quantized models and implementing EAGLE3 speculative decoding on NVIDIA GB300 clusters, validated across multiple benchmarks without quality degradation.
CoreWeave's Inference platform offers flexible deployment options including Serverless Inference, Dedicated Inference, and Inference on CoreWeave Kubernetes Service to allow customers to optimize cost-performance for different workloads.
This achievement demonstrates how infrastructure optimization across GPU selection, quantization, memory architecture, and kernel tuning directly impacts real-world inference performance for enterprise AI applications.