LongCat-Flash-Lite

LongCat-Flash-Lite is a lightweight MoE model from Meituan with 68.5B total parameters and only 2.9B-4.5B activated per token. It explores N-gram embedding expansion as a new scaling direction, supporting 256K context length via YaRN.

MATH-500

96.8%

i
MMLU

85.5%

i
CMMLU

82.5%

i
MMLU-Pro

78.3%

i
Tau2 Retail

73.1%

i
Tau2 Telecom

72.8%

i
AIME 2024

72.2%

i
GPQA

66.8%

i
AIME 2025

63.2%

i
Tau2 Airline

58.0%

i
SWE-Bench Verified

54.4%

i
SWE-bench Multilingual

38.1%

i
Terminal-Bench

33.8%

i