跳轉到主要內容

Kimi K2

KimiKimi開源權重MIT · 商用許可

描述

Kimi K2-Instruct-0905 is the latest, most capable version of Kimi K2, achieving state-of-the-art performance in frontier knowledge, math, and coding among non-thinking models. This Mixture-of-Experts model features 32 billion activated parameters and 1 trillion total parameters, meticulously optimized for agentic tasks. Key features include enhanced agentic coding intelligence, extended context length to 256K tokens, and a hybrid architecture trained with MuonClip optimizer on 15.5T tokens. The model achieves 65.8% on SWE-bench Verified (single attempt), 47.3% on SWE-bench Multilingual, and excels at tool use with 70.6% on Tau2-retail. It is a reflex-grade model without long thinking, designed to act and execute complex tasks seamlessly.

發布日期
2025-07-11
參數規模
1.0T
上下文長度
262K
支援模態
text

能力雷達圖

60
general
50
coding
70
reasoning
68
science估算
60
agents
0
multimodal

缺少專門科學評測時,Science 由 LLM Stats 科學得分或推理能力估算。

排行榜排名

暫無排名資料

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Chat

IFEvalGoogle Research (2023)89.8%自報
Tau2 Airline56.5%自報
Multi-Challenge54.1%自報

Communication

Tau2 Retail70.6%自報
Tau2 Telecom65.8%自報

Factuality

SimpleQA31.0%自報

General

C-Eval92.5%自報
MMLU89.5%自報
MultiPL-E85.7%自報
TriviaQA85.1%自報
CSimpleQA78.4%自報
ACEBench76.5%自報
Aider-Polyglot60.0%自報
SWE-bench Verified (Agentless)51.8%自報

Language

MMLU-Redux92.7%自報
MMLU-redux-2.090.2%自報
MMLU-Pro81.1%自報

Math

MATH-50097.4%自報
GSM8k97.3%自報
CBNSL95.6%自報
LiveBench76.4%自報
CNMO 202474.3%自報
MATH70.2%自報
AIME 202469.6%自報
PolyMath-en65.1%自報
AIME 202549.5%自報
HMMT 202538.8%自報

Reasoning

HumanEvalOpenAI (2021)93.3%自報
AutoLogi89.5%自報
ZebraLogic89.0%自報
HumanEval-ER81.1%自報
EvalPlus0.80 / 100自報
MuSR76.4%自報
GPQANYU + Cohere + Anthropic (2023)75.1%自報
SWE-bench Verified (Multiple Attempts)71.6%自報
SWE-bench Verified (Agentic Coding)65.8%自報
SWE-Bench Verified65.8%自報
SuperGPQA57.2%自報
LiveCodeBench v653.7%自報
LiveCodeBench53.7%自報
SWE-bench Multilingual47.3%自報
Terminal-Bench30.0%自報
OJBench27.1%自報
Terminus25.0%自報
Humanity's Last Exam4.7%自報

AA 評測指數

(Artificial Analysis)

暫無 AA 評測資料

LLM Stats 分類評分

(LLM Stats (zeroeval))
Instruction Following
90
Language
90
Structured Output
90
Legal
80
Finance
80
Healthcare
80
Biology
80
Chat
70
Math
70
Physics
70
Frontend Development
70
Chemistry
70
Reasoning
60
General
60
Communication
60
Economics
60
Tool Calling
60
Code
50
Factuality
30
Agents
20
Vision
0

定價

輸入價格$0.57 / 1M tokens
輸出價格$2.3 / 1M tokens
混合價格(3:1)$1.075 / 1M tokens

速度

Tokens/秒49.1
首Token延遲1.92s
首回答延遲1.92s

供應商價格排行

供應商價格排行

8 個供應商

最便宜: OpenCode最貴: AI/ML API
供應商輸入輸出
1OpenCode最便宜
$0.4
$2.5
2Kimi主要
$0.57
$2.3
3OpenRouter
$0.57
$2.3
4Kilo Gateway
$0.57
$2.3
5Vercel AI Gateway
$0.57
$2.3
6DevPass (LLM Gateway)
$0.57
$2.3
7Qiniu
$0.57971
$2.31884
8AI/ML API
$0.60306
$2.4334

比較該模型在不同 API 供應商之間的定價。

外部連結