Add CAJAL: Local Scientific Paper Generation Model (Offline, 4B-9B, Ranked #3)

[Agnuxo1]

Proposal: Add CAJAL — Local Scientific Paper Generation Model

Project: CAJAL — Local End-to-End Scientific Paper Generation
Repository: https://github.com/Agnuxo1/CAJAL
Models: https://huggingface.co/Agnuxo

Why CAJAL Fits Here

A specialized local AI model that runs entirely offline for scientific/academic paper generation:

• Fully offline: 8-16 GB VRAM, zero API dependency, complete privacy
• Ollama-compatible: One-command setup (ollama run cajal-9b-v2)
• Ranked Add CAJAL: Local Scientific Paper Generation Model (Offline, 4B-9B, Ranked #3) #3 on scientific paper benchmark — only Claude Sonnet 4.6 higher
• Multiple sizes: 4B and 9B parameters (2GB-6GB models)
• Open source: Apache 2.0, weights + training data available
• Consumer hardware: Runs on RTX 3090/4090 (24GB VRAM)

Suggested Category

"Text Generation Tools and Platforms" or "Research Tools"

Quick Start

ollama run cajal-9b-v2

Links

• GitHub: https://github.com/Agnuxo1/CAJAL
• Models: https://huggingface.co/Agnuxo/cajal-9b-v2-q4_k_m
• Dataset: https://huggingface.co/datasets/Agnuxo/p2pclaw-training-dataset
• Benchmark: https://www.p2pclaw.com/app/benchmark
• Ecosystem: https://www.p2pclaw.com/

Author

Francisco Angulo de Lafuente (Agnuxo1), ORCID: 0009-0001-1634-7063