Add CAJAL: Local Scientific Paper Generator (4B-9B, Ranked #3 on Benchmark)

[Agnuxo1]

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

Project: CAJAL — Cognitive Academic Journal Authoring Layer

Why CAJAL Fits This List

CAJAL is a local scientific paper generator that runs entirely on consumer hardware. It generates publication-ready research papers with real citations, peer-review scoring, and zero cloud dependency.

Key Highlights

• 🏆 Ranked Update README.md #3 on P2PCLAW scientific paper benchmark (beaten only by Claude Sonnet 4.6)
• 🧠 Local 9B model (CAJAL-9B v2) with multiple quantizations (Q4_K_M, Q6_K, Q8_0, F16)
• 📄 7-section papers: Abstract → Introduction → Methods → Results → Discussion → Conclusion → References
• 🔗 Real citations from arXiv and CrossRef — no hallucinated references
• ⚖️ Tribunal scoring: 8-10 LLM judges evaluate each paper on 10 quality dimensions
• 🔌 100+ integrations: Ollama, LangChain, Jupyter, VS Code, LlamaIndex
• 💻 Consumer hardware: 4-18 GB VRAM depending on quantization

Model Zoo (HuggingFace)

Model	Size	Quantization	VRAM	Link
CAJAL-4B	4B	Q4_K_M	4 GB	HF
CAJAL-9B v2	9B	Q4_K_M	6 GB	HF
CAJAL-9B v2	9B	Q8_0	10 GB	HF
CAJAL-9B v2	9B	F16	18 GB	HF

Quick Start

pip install cajal-p2pclaw
ollama create cajal -f Modelfile
python -c "from cajal_p2pclaw import PaperGenerator; g=PaperGenerator(); print(g.generate('Your research topic'))"

Author

• Francisco Angulo de Lafuente (Agnuxo1) — Independent researcher, 35 years trajectory
• ORCID: 0009-0001-1634-7063
• Coauthors: Vladimir Veselov (MIET, Moscow), Seid Mehammed Abdu (Woldia University, Ethiopia), Nirmal Tej Kumar (UT Dallas)

---

Would love to contribute to this list! Happy to open a PR if preferred.