Hijin Model Catalog
Official, verified neural packages engineered for local execution within the Hijin desktop workspace. Models are installed and managed directly through the application.
Managed In-App Experience
Hijin is designed around a curated, secure model distribution architecture. You do not need to download raw model files manually through your browser. Once the Hijin application is installed, the in-app Model Hub handles streaming, SHA-256 checksum verification, and activation.
Hijin Financial Specialist
Domain-adapted neural weights fine-tuned specifically for invoice line-item extraction, two-way and three-way purchase order reconciliation, and financial discrepancy detection.
Gemma 3 4B IT (Conversation Core)
Instruction-tuned foundational reasoning core developed by Google DeepMind. Provides conversational analysis, document query synthesis, and executive briefing memo drafting.
Model Package Integrity & Security Policy
To safeguard enterprise financial integrity, Hijin executes only officially verified, manifest-signed model packages. The application does not permit unauthenticated third-party model uploads, arbitrary web searches, or unverified community GGUF imports. Every model downloaded by the desktop client must pass SHA-256 checksum verification before activation.