跳轉到主要內容

LongCat-Flash-Thinking

MeituanOpen WeightMIT · Commercial OK

描述

LongCat-Flash-Thinking is Meituan's reasoning model built on the LongCat-Flash foundation with 560B total parameters (MoE, ~27B activated). It introduces a training pipeline specifically tuned for advanced reasoning, featuring Re-thinking Mode that delivers parallel reasoning paths for sophisticated decision-making. Achieves strong performance on mathematical reasoning, agentic tool use, and formal theorem proving benchmarks.

發布日期
2025-09-22
參數規模
560.0B
上下文長度
支援模態
text

能力雷達圖

80
general
70
coding
90
reasoning
68
science估算
70
agents
0
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8
74.0
LS
推理能力59
62.0
LS

基準測試分數 (LLM Stats)

Agents

BFCL-v374.4%自報

Biology

GPQA81.5%自報

Code

LiveCodeBench79.4%自報
SWE-Bench Verified59.4%自報

Communication

Tau2 Telecom83.1%自報
Tau2 Retail71.5%自報
Tau2 Airline67.5%自報

Finance

MMLU-Pro82.6%自報

General

MMLU-Redux89.3%自報

Math

MATH-50099.2%自報
AIME 202493.3%自報
AIME 202590.6%自報

Reasoning

ZebraLogic95.5%自報
ARC-AGI50.3%自報

AA 評測指數

暫無 AA 評測資料

LLM Stats 分類評分

Language
90
Math
90
Biology
80
Chemistry
80
Finance
80
General
80
Healthcare
80
Legal
80
Physics
80
Reasoning
80
Structured Output
70
Tool Calling
70
Agents
70
Code
70
Communication
70
Frontend Development
60
Spatial Reasoning
50
Vision
50

定價

暫無定價資料

速度

暫無速度資料

可用提供商

(LS 內部計價單位)

暫無提供商資料

外部連結