Kimi K2

Training compute

3×10²⁴ FLOP

Parameters

1T

Published

Jul 11, 2025

Kimi K2 is an AI model developed by Moonshot (China), first published in July 2025. It works in the language domain, on tasks such as language modeling/generation, code generation, question answering and 3 more.

Training it took an estimated 3×10²⁴ FLOP of compute (estimation method: operation counting). The model has 1,000,000,000,000 parameters. It was trained on roughly 15.5T datapoints. Training ran on NVIDIA H800 SXM5.

Access: Open weights (restricted use). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Moonshot
Country of organization: China
Domain: Language
Task: Language modeling/generation, Code generation, Question answering, Quantitative reasoning, Search, Table tasks
Training compute: 3×10²⁴ FLOP
Compute estimation method: Operation counting
Parameters: 1,000,000,000,000
Dataset size: 15.5T
Training hardware: NVIDIA H800 SXM5
Model accessibility: Open weights (restricted use)
Open weights: Yes
Epoch confidence: Confident

Benchmark results

01

Epoch Capabilities Index

ECI Score

140.4