Latest Content
Kimi K2 Thinking icon

Kimi K2 Thinking

NVIDIA
Kimi-K2-Thinking is a reasoning-focused Mixture-of-Experts (MoE) large language model, designed as a step-by-step thinking agent capable of dynamically invoking tools during reasoning. The model features a 1T parameter architecture with 32B activated parameters, 61 transformer layers, 64 attention heads, and 384 experts (8 experts per token). It supports a 256K token context window and is optimized for deep reasoning and long-horizon autonomous workflows. The model employs native INT4 quantization with Quantization-Aware Training (QAT) to reduce inference latency and memory usage while maintaining strong performance.
Type MoE LLM
CapabilitiesText Generation, Instruction Following, Reasoning, Mathematical Reasoning+5 more
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 --gpus all 
 --shm-size 128g 
 -p 8000:8000 
 -v ~/.cache/huggingface:/root/.cache/huggingface 
 -e HF_TOKEN='YOUR_HF_TOKEN' 
 --ipc=host 
 lmsysorg/sglang:v0.5.8-cu130 
 python3 -m sglang.launch_server 
 --model-path moonshotai/Kimi-K2-Thinking 
 --host 0.0.0.0 
 --port 8000 
 --max-running-requests 1024 
 --tp 8 
 --tool-call-parser kimi_k2 
 --reasoning-parser kimi_k2 
 --mem-fraction-static 0.90 
 --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