LongCat 2.5
Active Parameters
1.6T
Context Length
524K
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
228 KB/tok
License
MIT License
Release Date
4 Nov 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Self-hosted only
Hardware & System Requirements
Precision
Context
Workload Memory: 3073.92 GB
FP16 · 8K context
Base Weights: 3069.64 GB
KV Cache: 1.78 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
32x NVIDIA H200 141GB
3073.9 / 4060.8 GB usable · TP 8 · PP 4
Determine requirements for your specific workload:
Upgrade to Pro
Architecture Diagram
Evaluation Benchmarks
No evaluation benchmarks for LongCat 2.5 available.
Rankings
Overall Rank
-
Coding Rank
-
About LongCat 2.5
LongCat 2.5 is Meituan's next-generation open foundational language model optimized for extended context reasoning and high-throughput production workloads. It delivers enhanced instruction following and retrieval over extensive contexts.
Specifications
Attention
Attention Structure
DeepSeek Sparse Attention
Attention Heads
64
Attention Head Dimension
192
Position Embedding
ROPE
RoPE Theta
1,000,000
KV Latent Dimension
576
Indexer Key Dimension
128
Normalization
RMS Normalization
Dimensions
Base Precision
BF16
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
0
FFN Intermediate Size (per Expert)
2,048