跳轉到主要內容

Kimi K2.6 (Non-reasoning)

KimiKimiOpen WeightModified MIT License

描述

Kimi K2.6 is Moonshot AI's open-source, native multimodal agentic model focused on state-of-the-art coding, long-horizon execution, and agent swarm capabilities. It scales horizontally to 300 sub-agents executing 4,000 coordinated steps, dynamically decomposing tasks into parallel, domain-specialized subtasks. K2.6 unifies text, image, and video input with thinking and non-thinking modes, supports a 256K context, and powers proactive 24/7 background agents that manage schedules, execute code, and orchestrate cross-platform operations without human oversight.

發布日期
2026-04-20
參數規模
1.0T
上下文長度
262K
支援模態
image, text, video

能力雷達圖

38
general
39
coding
79
reasoning
53
science估算
60
agents
80
multimodal

Science 在缺少專門科學評測時使用推理能力代理估算。

排行榜排名

領域#排名分數來源
智能体与工具20
68.0
LS
代码能力榜69
67.0
AA
通用能力榜88
70.0
AA
多模态榜22
84.0
LS
推理能力62
61.0
LS
科学能力90
63.0
AA

基準測試分數 (LLM Stats)

Agents

BrowseComp86.3%自報
DeepSearchQA83.0%自報
Claw-Eval80.9%自報
WideSearch80.8%自報
OSWorld-Verified73.1%自報
Terminal-Bench 2.066.7%自報
SWE-Bench Pro58.6%自報
MCP-Mark55.9%自報
Toolathlon50.0%自報
APEX-Agents27.9%自報

Biology

GPQA90.5%自報
SciCode52.2%自報

Code

SWE-Bench Verified80.2%自報
SWE-bench Multilingual76.7%自報

General

LiveCodeBench v689.6%自報
MMMU-Pro80.1%自報

Math

AIME 202696.4%自報
MathVision93.2%自報
HMMT Feb 2692.7%自報
IMO-AnswerBench86.0%自報
Humanity's Last Exam36.4%自報

Multimodal

V*96.9%自報
CharXiv-R86.7%自報
BabyVision68.5%自報

Reasoning

OJBench60.6%自報

AA 評測指數

Intelligence Index
43.0
Coding Index
38.4
Tau2
0.9
Gpqa
0.8
Lcr
0.6
Ifbench
0.4
Scicode
0.4
Terminalbench Hard
0.4
Hle
0.2

LLM Stats 分類評分

Vision
80
Coding
80
Frontend Development
80
General
80
Math
80
Multimodal
80
Reasoning
80
Search
80
Agents
70
Biology
70
Chemistry
70
Code
70
Physics
70
Tool Calling
60

定價

輸入價格$0.95 / 1M tokens
輸出價格$4 / 1M tokens
混合價格(3:1)$1.712 / 1M tokens

速度

Tokens/秒28.5 tokens/s
首Token延遲1.71s
首回答延遲1.71s

可用提供商

(LS 內部計價單位)
提供商輸入價格輸出價格
Novita950K4.0M
Moonshot AI950K4.0M

外部連結