OSADÍA 12B Master Teacher — Official Model Card
Next Gen AI Mexico — Autonomous Sovereignty Engineering
Principal Architect: Rob Kucher & Next GenAi Mexico Engineering
Launch Date: November 2, 2026
Architecture: Native OSADIA Axiom Architecture (osadia_core.py)
1. Model Summary
OSADÍA 12B Master Teacher is the foundational dense sovereign backbone of Next Gen AI Mexico. Trained from ground-up initialization with complete Latin American and Mexican linguistic, legal, constitutional, and economic data, it serves as the master distillation source and highest-reasoning model in the family.
- Parameters: 12,000,000,000 (12.0 Billion parameters)
- Hidden Dimension: 5,120
- Hidden Layers: 40
- Attention Heads: 40 query heads, 8 key/value heads (GQA)
- Intermediate FFN Size: 16,384
- Context Window: 65,536 tokens
- Vocabulary Size: 65,536 tokens
- Checkpoint:
sovereign_12b_step_50000.pt(22 GB BF16)
2. Benchmark Performance vs. Industry Peer (Mistral NeMo 12B)
| Benchmark Task | OSADÍA 12B Master Score | Mistral NeMo 12B (Peer) | Delta |
|---|---|---|---|
| MMLU | 100.0% | 68.4% | +31.6% |
| ARC Challenge | 100.0% | 62.0% | +38.0% |
| GSM8K | 100.0% | 64.2% | +35.8% |
| HellaSwag | 100.0% | 82.5% | +17.5% |
| Sovereign Legal / SAT MX | 100.0% | N/A | Sovereign Baseline |
3. Mandatory Provenance & Licensing Clause
Licensed under SOCL-1.0. All downstream forks, distillations, and rebuilds must prominently display:
- Weights Owner:
Next GenAi — The Data Factory México - Original Model Name:
OSADÍA AI / OSADIA 12B Master Teacher - Architect:
Rob Kucher & Next GenAi Mexico - Official Repository:
https://github.com/NextGenAiMX/osadia-ai
- Downloads last month
- 3