The intuyt Model Family by LLM-MX

The intuyt family represents a sovereign, highly efficient suite of Large Language Models developed under the LLM-MX initiative. Engineered to deeply understand Mexican Spanish and optimized for GreenAGI principles, the models span from ultra-efficient edge computing prototypes to robust reasoning engines.

🚀 Model Lineup

1. intuyt:nano (1.7B VLM) - Available Now

• Base Architecture: Qwen3 (1.7B)
• Tuning Method: Extensive Fine-Tuning
• Quantization: Q4_K_M (1.1GB)
• Role: Edge computing optimized. Ideal for low-latency, offline local inference on resource-constrained devices (tablets, local gateways) requiring Vision-Language (VLM) capabilities and native Mexican Spanish.

2. intuyt:micro (4B) - Prototyping Phase

• Base Architecture: Qwen3:4b-instruct
• Tuning Method: Heavy System Prompting (Giga-Prompting)
• Role: Advanced logic and agentic workflow orchestration.
• Technical Justification for Giga-Prompting: While traditional fine-tuning permanently updates model weights, intuyt:micro is currently deployed using Heavy System Prompting (Giga-Prompting). In modern, highly-aligned instruct architectures like Qwen3, comprehensive system prompting is a scientifically valid and compute-efficient method for model tuning and rapid prototyping. By front-loading strict behavioral constraints, topological schemas, and few-shot reasoning examples into an expanded context window, we effectively perform activation steering—guiding the model’s latent representations toward the desired outputs without the risk of catastrophic forgetting. This approach allows for rapid, cost-effective iteration of complex agent behaviors (e.g., formatting for n8n or OpenClaw pipelines). Furthermore, it acts as a synthetic data engine, generating the exact, highly-curated conversational datasets required for its upcoming explicit fine-tuning phase, perfectly aligning with GreenAGI principles by saving massive amounts of preliminary compute.

3. intuyt:macro (30B) - Upcoming

• Base Architecture: Qwen3:30b-a3b-instruct (Mixture of Experts)
• Role: The flagship reasoning engine of the LLM-MX initiative. Designed for enterprise-grade deployments, massive transdisciplinary data analysis, and serving as the central orchestrator for complex GreenAGI multi-agent topologies.

🤝 Ecosystem & Acknowledgments

The development of the intuyt family is driven by a robust network of academic, research, and industry partners:

Core Research & Applied AI

• intu.yt: The main hub for the LLM-MX initiative and tailored Mexican AI models.
• TRAI-L: The Transdisciplinary Research for Augmented Innovation Laboratory, leading the theoretical frameworks for GreenAGI and modular neurosymbolic AI.
• ArtInt.Tech: Bridging the gap between cutting-edge local LLM research and enterprise agentic workflow implementations.

Academic & Institutional Partners

• NLP Lab @ CIC-IPN: The Natural Language Processing Laboratory at the Centro de Investigación en Computación (IPN), driving foundational research in computational linguistics and indigenous language preservation.
• IPICYT: Instituto Potosino de Investigación Científica y Tecnológica, fostering advanced scientific research and technological development in Mexico.
• Universidad de Sonora (Unison) & Summa IT: Key collaborators in the expansion of AI capabilities, technological integration, and applied research across academic and enterprise sectors.

Hardware & Infrastructure

• AMD: Providing the high-performance ROCm compute power (Instinct accelerators) required for local deployment and fine-tuning.
• Dell Technologies: Supplying the enterprise-grade server architecture that supports our intensive local AI deployments and research pipelines.

🙌 Credits

The lead researcher on this model version is Ana Gabriela Ordoñez-Güemes. Our deepest gratitude to her unwavering effort.