LongCat-Flash-Thinking

Name: LongCat-Flash-Thinking
Author: Meituan

Meituan

LongCat Flash Thinking is a reasoning-enhanced variant of Meituan's LongCat model, featuring extended chain-of-thought capabilities for complex problem solving. It combines long-context processing with deep reasoning while maintaining fast inference speed.

Key Specifications

Parameters

560.0B

Context

Release Date

September 21, 2025

Average Score

79.8%

API Documentation Research Paper Repository Model Weights

Timeline

Key dates in the model's history

Announcement

September 21, 2025

Last Update

February 17, 2026

Today

September 10, 2026

Technical Specifications

Parameters

560.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

• Self-reported

59.4%

Reasoning

Logical reasoning and analysis

GPQA

• Self-reported

81.5%

Other Tests

Specialized benchmarks

MATH-500

Mean@1 • Self-reported

99.0%

ZebraLogic

Mean@1 • Self-reported

95.5%

AIME 2024

Mean@32 • Self-reported

93.0%

AIME 2025

Mean@32 • Self-reported

91.0%

MMLU-Redux

• Self-reported

89.0%

Tau2 Telecom

Mean@4 • Self-reported

83.0%

MMLU-Pro

• Self-reported

82.6%

Tau2 Retail

Mean@4 • Self-reported

71.5%

ARC-AGI

Mean@1 • Self-reported

50.3%

BFCL-v3

Self-reported by the model provider • Self-reported

74.4%

LiveCodeBench

Mean@4 • Self-reported

79.4%

Tau2 Airline

Mean@4 • Self-reported

67.5%

License & Metadata

License

mit

Announcement Date

September 21, 2025

Last Updated

February 17, 2026

Similar Models

All Models