About
Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It is optimized for...
Related Models
Pricing
Input
$0.57
per 1M tokens
Output
$2.30
per 1M tokens
Blended
$1.00
per 1M tokens
Cheaper than 35% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$1.00
Monthly
$30.07
vs. Similar Models
OpenAI: GPT-4.1Q:0.0
$3.50+249%
inclusionAI: Ling-2.6-flashQ:-0.1
$0.01-99%
GLM-4.5 (Reasoning)Q:+0.1
$1.00-0%
Qwen3 Max (Preview)Q:-0.2
$2.40+139%
Performance
26
tokens/sec
Faster than 1% of models
1.51
seconds
Faster than 34% of models
1.51
seconds
Faster than 57% of models
Market Median
94 tok/s
72% slower
Median TTFT
1.10s
36% slower
Throughput/Dollar
26
tok/s per $/1M
Speed Comparison
Qwen3.5 2B (Non-reasoning)
27 tok/s+1%
Gemma 3 12B Instruct
26 tok/s-1%
MoonshotAI: Kimi K2 0905
26 tok/s-3%
Context Window
131K
tokens
Larger than 27% of models
Max Output
33K
tokens
25% of context
Benchmarks
MMLU-Pro
82.4%
GPQA Diamond
76.6%
HLE
7.0%
LiveCodeBench
55.6%
SciCode
34.5%
TerminalBench Hard
15.9%
MATH-500
97.1%
AIME
69.3%
AIME 2025
57.0%
IFBench
41.5%
Long Context Recall
51.0%
Tau2
61.1%
Market AverageTop Score