Meituan
Meituan: LongCat 2.0
meituan/longcat-2.0
LongCat 2.0 is a sparse mixture-of-experts language model from Meituan, with 48B active parameters out of 1.6T total. It is suited for coding, repository-level changes, long-horizon problem solving, and agentic workflows.
Input
$0.3
/ 1M tokens
Output
$1.2
/ 1M tokens
Context window
1.05M
tokens
Limits
Context
1,048,756
tokens
Max output
262,144
25% of the context window
Parameters
1,775,560,491,136
Model facts
Author
Meituan
Tokenizer
Other
Modality
Text -> Text
License
MIT
Created
Jul 20, 2026
Hugging Face
meituan-longcat/LongCat-2.0
Publication status
Hugging Face: verifiedLicense: publishedParameter count: published
Modalities
Input · TextOutput · Text
Generation controls
Frequency penaltyInclude reasoningLogit biasMax tokensMin pPresence penaltyReasoningRepetition penaltySeedStopTemperatureTool choiceToolsTop kTop p
Capability scores
14
Agentic index
Better than 37% of models compared
45.3
Coding index
Better than 41% of models compared
19.1
Intelligence index
Better than 45% of models compared