AI MODEL STACK
Four specialized AI models. One sovereign hardware stack. Zero cloud dependency.
Every Digital Sovereignty deployment runs four purpose-selected local models. Each model is optimized for a specific operational role. Together they form a complete autonomous monitoring and intelligence layer that never sends data outside your infrastructure.
Google's multimodal model handles visual monitoring tasks. It screenshots and analyzes product pages, detecting layout breaks, missing content, and visual anomalies that text-only models miss.
Specs
Q4 quantization · ~8GB RAM active · 12B parameters
Best for
Microsoft's compact reasoning model runs rapid pattern analysis on logs, metrics, and system events. Its small footprint means it can run alongside other models without straining hardware.
Specs
Q4 quantization · ~2.5GB RAM active · 3.8B parameters
Best for
Alibaba's code-focused model handles technical integrity checks. It verifies HTML structure, catches broken links in code, and performs logical consistency checks across monitored systems.
Specs
Q4 quantization · ~3.4GB RAM active · 4B parameters
Best for
Meta's versatile model generates reports, briefings, and localized content. It handles Czech and English fluently, making it ideal for producing morning briefs and operational summaries for local teams.
Specs
Q4_K_M quantization · ~4.9GB RAM active · 8B parameters · prompt 34.7 t/s · generation 8.4 t/s
Best for
COMPARISON
| Feature | Gemma 3 12B | Phi-4 mini 3.8B | Qwen 3.5 4B | Llama 3.1 8B |
|---|---|---|---|---|
| Role | Vision & Monitoring | Anomaly Detection | Code & Logic | Content & Localization |
| Size | 8.1 GB | 2.5 GB | 3.4 GB | 4.9 GB |
| Parameters | 12B | 3.8B | 4B | 8B |
| Quantization | Q4 | Q4 | Q4 | Q4_K_M |
| RAM (active) | ~8 GB | ~2.5 GB | ~3.4 GB | ~4.9 GB |
| Multimodal | Yes (vision) | No | No | No |
| Best Speed | Medium | Fast | Fast | Medium |
| Languages | Multi | Multi | Code-focused | Czech + English |
| Total Stack RAM | ~18.9 GB combined. Fits in 32GB with headroom. | |||
WHY LOCAL, NOT CLOUD
Running models locally means your operational data never leaves your infrastructure. No API costs. No rate limits. No vendor lock-in. The EU AI Act Article 14 requires human oversight of AI systems. Local models make this possible because you control exactly what they see and how they respond.
Digital Sovereignty provides independent operator capability assessments and infrastructure. The Proof of Operator seal constitutes evidence that an assessment was conducted by Brainiac Ltd. It is not a legal certification of compliance with EU Regulation 2024/1689 or any other regulation. Brainiac Ltd is not liable for regulatory decisions based on the seal. Companies should consult their own legal counsel.