modelgrep
M

MoonshotAI: Kimi K2 Thinking

moonshotai/kimi-k2-thinking

104th smartest of 178ReasoningToolsJSON
Use via OpenRouter ↗
Intelligence
24.1
104th of 178
Design Elo
1158
Website
Speed
57
144th fastest
Latency
558ms
first token
Input price
$0.600
173rd cheapest
Context
262K
262K max out

How it compares

Smarter than42%
of all ranked models
Faster than51%
of all ranked models
Cheaper than42%
of all ranked models

Overview

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...

Benchmarks

independent · via OpenRouter
Artificial Analysis40th percentile
Intelligence Index
24.1
Coding Index
15.5
Agentic Index
15.3
GPQA Diamond
71%
Humanity's Last Exam
10%
SciCode
33%
Tau²-Bench (agentic)
25%
Design Arena · Elo503 tournaments
Website
1158

Providers & pricing (3)

ProviderIn $/MOut $/MUptime
Google$0.600$2.5098.6%
AtlasCloudint4$0.600$2.50
Novitabf16$0.600$2.50

Specifications

Context window262K
Max output262K
Knowledge cutoff
Input modalitiestext
Output modalitiestext
Prompt caching
Cache read price
ModeratedNo

Kimi K2 Thinking FAQ

How much does Kimi K2 Thinking cost?

Kimi K2 Thinking costs $0.600 per million input tokens and $2.50 per million output tokens via OpenRouter, making it 173rd cheapest of 298 paid models.

How smart is Kimi K2 Thinking?

Kimi K2 Thinking scores 24.1 on the Artificial Analysis Intelligence Index, ranking 104th of 178 benchmarked models, with a GPQA Diamond score of 71%.

How fast is Kimi K2 Thinking?

Kimi K2 Thinking generates around 57 tokens per second with 558ms time-to-first-token (p50), the 144th fastest tracked model.

What is Kimi K2 Thinking's context window?

Kimi K2 Thinking supports a 262K-token context window and can output up to 262K tokens. It accepts text input.

Compare head-to-head