ERNIE-4.5-300B-A47B-Base
Active Parameters
300B
Context Length
131K
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
324 KB/tok
License
Apache 2.0
Release Date
30 Jun 2025
Knowledge Cutoff
Jun 2025
API Pricing (per 1M)
Input: $0.28 · Output: $1.10
Hardware Sizing
Precision
Context
Workload Memory: 651.37 GB
FP16 · 8K context
Base Weights: 646.34 GB
KV Cache: 2.53 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
8x NVIDIA H200 141GB
651.4 / 1015.2 GB usable · TP 8
Determine requirements for datacenter hardware:
Upgrade to Pro
Architecture Diagram
Evaluation Benchmarks
No evaluation benchmarks for ERNIE-4.5-300B-A47B-Base available.
Rankings
Overall Rank
-
Coding Rank
-
About ERNIE-4.5-300B-A47B-Base
ERNIE-4.5-300B-Base is a 300B text Mixture-of-Experts base model from Baidu activating 47B parameters for large-scale bilingual pretraining. It incorporates intra-node expert parallelism and Grouped-Query Attention across a 131K context window.
Technical Specifications
Attention
Attention Structure
Grouped-Query Attention
Attention Heads
64
Key-Value Heads
8
Position Embedding
Absolute Position Embedding
RoPE Theta
500,000
Normalization
Layer Normalization
Activation Function
GELU
Dimensions
Hidden Dimension Size
12,288
Number of Layers
54
FFN Intermediate Size (Dense)
3,584
Tokenizer
Vocabulary Size
103,424
Mixture of Experts
Total Expert Parameters
47.0B
Number of Experts
64
Active Experts
8
Shared Experts
0
FFN Intermediate Size (per Expert)
3,584
Dense Layers Before MoE
3
About ERNIE 4.5
The Baidu ERNIE 4.5 family consists of ten large-scale multimodal models. They utilize a heterogeneous Mixture-of-Experts (MoE) architecture, which enables parameter sharing across modalities while also employing dedicated parameters for specific modalities, supporting efficient language and multimodal processing.