# TinyMind GGUF-X

GGUF-X is a TinyMind runtime package: a GGUF v3-compatible model plus sidecar evidence, decode policy, and lineage gates.
It does not mutate GGUF tensors directly and does not claim better weights without a real merge/export/eval report.

## Files

- Modelfile: `reports\ggufx_purecode_latest\Modelfile.evo`
- Spec: `reports\ggufx_purecode_latest\tinymind_ggufx_spec.json`
- Eval prompts: `reports\ggufx_purecode_latest\gguf_evo_eval_prompts.jsonl`
- Create script: `reports\ggufx_purecode_latest\create_ollama_evo.ps1`
- Compare script: `reports\ggufx_purecode_latest\compare_rawzero_vs_evo.ps1`

## Claim Boundary

- Custom TinyMind format pack: allowed
- GGUF v3 runtime compatibility: allowed
- Better-than-source GGUF weights: blocked until same-prompt baseline evidence exists
- World-best/runtime superiority: blocked until external benchmark evidence exists