MaLLaM-3B
Parameters
3B
Context Length
4K
Modality
Text
Architecture
Dense
License
Apache-2.0
Release Date
15 Jan 2024
Knowledge Cutoff
Jan 2024
API Pricing (per 1M)
Self-hosted only
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
1x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
4,096 tokens
Consumer
1x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
No evaluation benchmarks for MaLLaM-3B available.
Rankings
Overall Rank
-
Coding Rank
-
About MaLLaM-3B
MaLLaM-3B is a sovereign 3B parameter foundation model developed by Malaysia AI and Mesolitica for Bahasa Malaysia and English applications. Trained on 90B tokens of local digital artifacts, it is optimized for low-latency edge deployment.
Technical Specifications
Attention
Attention Structure
Multi-Head Attention
Attention Heads
32
Key-Value Heads
8
Attention Head Dimension
100
Position Embedding
Absolute Position Embedding
RoPE Theta
10,000
Sliding Window Attention
Yes
Sliding Window Size
4,096
Sliding Window Ratio
-
Linear Attention
-
Linear Attention Ratio
-
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
3,200
Number of Layers
26
FFN Intermediate Size (Dense)
8,640
Multi-Token Prediction Heads
-
Tokenizer
Vocabulary Size
32,000
About MaLLaM
Malaysian Large Language Model (MaLLaM) is an open-source language model family developed to support Bahasa Malaysia and English. The model is trained on Malaysian text data including local news, literature, and digital content. It is designed to process Malaysian linguistic nuances and cultural context, available in multiple parameter sizes for different hardware deployments.
Other MaLLaM Models
- No related models available