BREAKING
83 LLM Architectures in One Gallery
0
models cataloged
0B
DeepSeek V3 params
0k
Gemma 3 context
Each Model Gets a Fact Sheet
Diagram
Decoder type
Attention mechanism
Layer mix
Fact sheet
Parameter count
Context length
License and date
Tracing the Design Lineage
1Dense models
2MoE designs
3MHA to GQA to MLA
Open Tools and Metadata
On the site
Two-model comparison
Printable posters
Open data
Public GitHub repo
Verify and reuse figures
A Curated Map of LLM Design
AI NEWS BLITZ
Sebastian Raschka has published a visual gallery tracing 83 large language model architectures.