The model registry starts as a source-backed template system. It should not publish broad comparative claims until the source trail, update date, and reviewer status are clear.
| Field | Purpose |
|---|---|
| Model identity | Name, publisher, release date, license, source links, and version notes. |
| Modalities | Text, image, video, audio, document, UI, spatial, temporal, or other supported inputs/outputs. |
| Architecture | High-level design, training approach, context handling, mixture-of-experts notes, OCR strategy, or memory design. |
| Evidence | Benchmarks, papers, repositories, demos, contamination notes, and last-reviewed date. |
| Deployment | Hardware, latency, memory, API availability, edge/cloud fit, and operational limits. |
| Interoperability | Known agent, tool, protocol, schema, or UAI-related implementation notes. |
Pilot queue
Document understanding
OCR-free and document-oriented LMMs need page templates that separate paper claims from deployment behavior.
UI and agent workflows
Grounded UI understanding, screenshot QA, and WebUI-to-code systems need reproducible examples and source links.
Long-context video
Temporal reasoning and memory-augmented video models need careful evaluation notes and modality labels.