# Model card

## Identity and lineage

- Base model, tokenizer, template, and immutable versions:
- Training/data lineage references:
- Adapters, merges, quantization, and serving transforms:

## Intended use

- Supported tasks and users:
- Explicitly unsupported or prohibited uses:
- Required surrounding controls:

## Evaluation evidence

- Evaluation contract and dataset versions:
- Overall, slice, calibration, robustness, and safety results:
- Hardware, decoding configuration, and uncertainty:

## Limitations and risks

- Known failure modes:
- Context, language, domain, and modality limits:
- Security/privacy considerations:

## Release decision

- Decision, owner, and date:
- Residual risks accepted by:
- Monitoring and rollback trigger:
- Next recertification date: