ERNIE-4.5-VL-28B-A3B-Base
Active Parameters
28B
Context Length
131K
Modality
Multimodal
Auxiliary Parameters
600M
Architecture
Mixture of Experts
KV Cache Rate
56 KB/tok
License
Apache 2.0
Release Date
30 Jun 2025
Knowledge Cutoff
Nov 2024
API Pricing (per 1M)
Input: $0.14 · Output: $0.56
Hardware Sizing
Precision
Context
Workload Memory: 17.37 GB
INT4 · 8K context
Base Weights: 14.44 GB
KV Cache: 0.44 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA L4 24GB
17.4 / 21.6 GB usable
2x RTX 4090 24GB
17.4 / 43.2 GB usable · TP 2
Apple M4 24GB
17.4 / 18.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Architecture Diagram
Evaluation Benchmarks
No evaluation benchmarks for ERNIE-4.5-VL-28B-A3B-Base available.
Rankings
Overall Rank
-
Coding Rank
-
About ERNIE-4.5-VL-28B-A3B-Base
ERNIE-4.5-VL-28B-Base is an open-weights multimodal Mixture-of-Experts base model from Baidu activating 3B parameters for vision-language tasks. It features heterogeneous MoE routing and 2D RoPE position embeddings over a 131K context window.
Technical Specifications
Attention
Attention Structure
Grouped-Query Attention
Attention Heads
20
Key-Value Heads
4
Position Embedding
Absolute Position Embedding
RoPE Theta
500,000
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Auxiliary Parameters
600M
Hidden Dimension Size
2,560
Number of Layers
28
FFN Intermediate Size (Dense)
12,288
Tokenizer
Vocabulary Size
103,424
Mixture of Experts
Total Expert Parameters
3.0B
Number of Experts
130
Active Experts
14
Shared Experts
2
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.