LongCat 2.0
Active Parameters
1.6T
Context Length
1.05M
Modality
Text
Architecture
Mixture of Experts (MoE)
License
MIT License
Release Date
5 Jul 2026
Knowledge Cutoff
-
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
251x RTX 4090
24GB VRAM
Datacenter
58x NVIDIA A100
80GB VRAM
Apple Silicon
56x Apple M3 Max
128GB VRAM
1,048,756 tokens
Consumer
268x RTX 4090
24GB VRAM
Datacenter
62x NVIDIA A100
80GB VRAM
Apple Silicon
60x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
No evaluation benchmarks for LongCat 2.0 available.
Rankings
Overall Rank
-
Coding Rank
-
About LongCat 2.0
LongCat 2.0 is a large-scale mixture-of-experts model from Meituan that activates 48B parameters out of 1.6T total. It is built for complex repository-level software modifications, extended reasoning, and agentic workflows.
Technical Specifications
Attention
Attention Structure
DeepSeek Sparse Attention
Attention Heads
64
Key-Value Heads
-
Attention Head Dimension
128
Position Embedding
ROPE
RoPE Theta
1,000,000
Sliding Window Attention
No
Sliding Window Size
-
Sliding Window Ratio
-
Linear Attention
No
Linear Attention Ratio
-
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
8,192
Number of Layers
38
FFN Intermediate Size (Dense)
12,288
Multi-Token Prediction Heads
3
Tokenizer
Vocabulary Size
163,840
Mixture of Experts
Total Expert Parameters
48.0B
Number of Experts
768
Active Experts
12
Shared Experts
-
FFN Intermediate Size (per Expert)
2,048
Dense Layers Before MoE
-
Resources
About LongCat
The LongCat model family developed by Meituan.
Other LongCat Models
- No related models available