# QFF — Quant Fidelity Frontier

> Publisher-reported LLM quantization claims, not independent QFF measurements.

Only compare compatible measurement sets. The all/overview set mixes methods. File size is not total VRAM. Preserve source URLs, capture dates, source revision pin status, units, size basis, methodology and copied-claim flags. Missing metrics are unknown, not zero.

## Data and guidance

- [Model manifest](https://delirious6423.github.io/quant-fidelity-frontier/models.json)
- [Interpretation and citation](https://delirious6423.github.io/quant-fidelity-frontier/guides/)
- [MCP source and setup](https://github.com/delirious6423/quant-fidelity-frontier/tree/main/qff_mcp)
- [Qwen/Qwen3.8-Flash-Next](https://delirious6423.github.io/quant-fidelity-frontier/qwen3.8-flash-next/data.json) — captured 2026-10-17T08:44:29.662Z
- [Qwen/Qwen3.8-27B](https://delirious6423.github.io/quant-fidelity-frontier/qwen3.8-27b/data.json) — captured 2026-10-17T19:29:59Z