Kimi K2 Thinking

Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning. Built on the trillion-parameter Mixture-of-Experts (MoE) architecture introduced in Kimi K2, it activates 32 billion parameters per forward pass and supports 256 k-token context windows. The model is optimized for persistent step-by-step thought, dynamic tool invocation, and complex reasoning workflows that span hundreds of turns. It interleaves step-by-step reasoning with tool use, enabling autonomous research, coding, and writing that can persist for hundreds of sequential actions without drift. It sets new open-source benchmarks on HLE, BrowseComp, SWE-Multilingual, and LiveCodeBench, while maintaining stable multi-agent behavior through 200–300 tool calls. Built on a large-scale MoE architecture with MuonClip optimization, it combines strong reasoning depth with high inference efficiency for demanding agentic and analytical tasks.

Licence

other

Architecture

DeepseekV3ForCausalLM

Parameters

1.06T

Context

262.1K tokens

Artifact format

safetensors

Quantization

compressed-tensors 4-bit

Download size

553 GB

Hardware to run it (inference, estimated)

592 GB VRAM minimum · 740 GB recommended · 740 GB system RAM · 554 GB storage

Estimated from parameter count and stored precision; verify against the selected runtime and context length.

Benchmarks

No benchmark observation recorded for this model.