Agent skill · writing content · aradotso
llm-architecture-gallery
llm-architecture-gallery — an agent skill for coding agents.
Why this skill is useful
Provides specific schema and conventions for contributing to the LLM Architecture Gallery, which the AI wouldn't reliably generate on its own.
What it needs
About 5k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
llm-architecture-gallery/ ├── models.yml # Primary data file — all model metadata └── images/ └── architectures/ # .webp architecture diagrams └── hero/ # Hero/banner images yaml GPT-4o: image: "/llm-architecture-gallery/images/architectures/gpt-4o.webp" date: '2024-05-13' yaml DeepSeek R1: image: "/llm-architecture-gallery/images/architectures/deepseek-v3-r1-671-billion.webp" date: '2025-01-20' parameters: "671B" architecture: "Mixture of Experts (MoE)" contextlength: "128K" organization: "DeepSeek" license: "MIT" paper: "https://arxiv.org/abs/2501.12948" code: "https://github.com/deepseek-ai/DeepSeek-R1" notes: "Reasoning model trained via reinforcement learning." Stored path in models.yml: /llm-architecture-gallery/images/architectures/deepseek-v3-r1-671-billion.webp Full public URL: https://sebastianraschka.com/llm-architecture-gallery/images/architectures/deepseek-v3-r1-671-billion.webp llama-3-70-billion.webp mistral-7b.webp deepseek-v3-r1-671-billion.webp gemma-2-27b.webp images/architectures/<model-slug>.webp yaml Llama 3.1 405B: image: "/llm-architecture-gallery/images/architectures/llama-3-1-405-billion.webp" date: '2024-07-23' parameters: "405B" architecture: "Transformer (GQA)" contextlength: "128K" organization: "Meta" license: "Llama 3.1 Community License" paper: "https://arxiv.org/abs/2407.21783" code: "https://github.com/meta-llama/llama3" notes: "Flagship open-weights model from Meta's Llama 3.1 family." bash Python — validate the file parses without errors python3 -c " import yaml, sys with open('models.yml') as f: data = yaml.safeload(f) print(f'OK: {len(data)} models loaded') for name, meta in data.items(): if 'image' not in meta: print(f'WARNING: {name!r} missing image field') if 'date' not in meta: print(f'WARNING: {name!r} missing date field') " bash git checkout -b add-llama-3-1-405b git add models.yml images/architectures/llama-3-1-405-billion.webp git commit -m "Add Llama 3.1 405B entry" git push origin add-llama-3-1-405b Then open PR at: h …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/llm-architecture-gallery