ai.sploit.dk

Model Registry Coding Agents System Cards

Registry — LLM / SLM / Domain-specialized

AI Model Registry

A structured index of language models — frontier LLMs, small language models, and domain-specialized systems for cybersecurity, medicine, finance and European languages. Every entry carries the same base record: vendor, country of origin, license, release, parameters and modalities. Specialized entries are further classified by how they specialize: purpose-built, domain fine-tune, or access-tier variant.

All models with disclosed parameter counts, on one log scale

Domain

Specialization

License

Type

Deployment

Data last updated August 2026. Verify parameters, licenses and release dates against vendor sources before relying on them for procurement or governance decisions.

Specialization taxonomy: purpose-built models are trained for their domain from the start; domain fine-tunes adapt a general base model; access-tier variants (dashed chip) are unchanged general frontier models offered with adjusted safeguards or access. The distinction matters for model selection: purpose-built and fine-tuned models are often small enough to run on-premises — relevant where cost per task and data privacy drive the choice — while an access-tier variant is the same hosted frontier model behind a different gate.

Token prices are USD list prices per 1M tokens on the vendor's own API unless the basis field says otherwise, and they change often — every price carries an as-of stamp, so treat anything older than a quarter as stale. Open-weight models are shown as self-hosted (infrastructure cost) rather than forced into a token price.