vLLM ships day-0 support for Kimi K3.1 Pro
vLLM announced day-0 support for Kimi K3.1 Pro, including sparse attention on a KDA hybrid architecture, a packed KV cache with FP8 or NVFP4 latents and MXFP4 index keys, and token-sharded prefill.
vLLM announced day-0 support for Kimi K3.1 Pro, including sparse attention on a KDA hybrid architecture, a packed KV cache with FP8 or NVFP4 latents and MXFP4 index keys, and token-sharded prefill.