Clef and Clef Flash: open-weights models for fast decisions
Open-weights models Clef and Clef Flash output probabilities and constrained classes instead of free-form text. Clef Flash, at roughly 9B parameters, targets sub-100ms latency and image input for real-time classification and triage.
- Clef Flash has about 9B parameters and targets low latency
- Models return probabilities and classes instead of free text
- Image support enables UI QA and document triage
- Open weights allow self-hosting and local deployment
Read next
AI