Latest Content
Kimi K2.5 icon

Kimi K2.5

NVIDIA
Kimi-K2.5 is a native multimodal Mixture-of-Experts (MoE) large language model designed for advanced coding, vision reasoning, and autonomous agentic workflows. The model features a 1T parameter architecture with 32B activated parameters, 61 layers, 64 attention heads, and 384 experts (8 experts per token). It supports up to a 256K token context window and integrates a 400M-parameter MoonViT vision encoder for cross-modal understanding. The model employs native INT4 quantization with Quantization-Aware Training (QAT) to reduce inference latency and memory usage while maintaining strong performance.
TypeVision-Language Model
CapabilitiesText Generation, Instruction Following, Reasoning, Mathematical Reasoning+6 more
Group Release DateJanuary 26, 2026
Links
Paper|Blog|HF Model Card
LicenseModified MIT

Inference Instructions

Deploy and run this model on NVIDIA B200 GPUs using the command below. Copy the command to get started with inference.

CONSOLE
docker run -it --rm 
 --runtime=nvidia 
 --gpus all 
 --ipc=host 
 --shm-size=128g 
 -p 8000:8000 
 -v ~/.cache/huggingface:/root/.cache/huggingface 
 -e HF_TOKEN='YOUR_HF_TOKEN' 
 -e LD_LIBRARY_PATH='/usr/local/nvidia/lib64:/usr/local/nvidia/lib:/usr/lib/x86_64-linux-gnu' 
 vllm/vllm-openai:v0.15.0-cu130 
 moonshotai/Kimi-K2.5 
  --tensor-parallel-size 8 
 --mm-encoder-tp-mode data 
 --max-model-len auto 
  --gpu-memory-utilization 0.90 
 --tool-call-parser kimi_k2 
 --reasoning-parser kimi_k2 
 --enable-auto-tool-choice 
 --max-num-seqs 1024 
 --disable-log-requests 
 --trust-remote-code

Model Benchmarks

Each model was tested with a fixed input size and total token volume while increasing concurrency to measure serving performance under load.

ITL vs Concurrency

Time to First Token

Throughput Scaling

Total Tokens/sec vs Avg TTFT

Vultr Cloud GPU

NVIDIA HGX B200

Deploy NVIDIA B200 on Vultr Cloud GPU

Deploy NVIDIA B200 instancesContact Sales