Overall score vs GGUF size
Axes are fitted to the shown results; near-identical file sizes are offset for legibility. Markers show published placement, R is the reference, and orange outlines mark the efficiency frontier.
Public DashboardFluid Bench · public results
Browse verified model evidence and published competition result sets from one sealed Fluid epoch.
TPN-005 · 4 runs
Published placement stays tied to its result set; selecting another concern changes the sort, not the placement.
| Compare | Placement | Model | Overall | GGUF size | Format | Details |
|---|---|---|---|---|---|---|
| 1 | 73.93% | 7.24 GB | GGUF | |||
| 2 | 73.79% | 7.23 GB | GGUF | |||
| 3 | 73.57% | 7.23 GB | GGUF | |||
| REF | 66.57% | 16.99 GB | GGUF F16 |
Model insight
TPN-005 · verified GGUF artifacts only.
Axes are fitted to the shown results; near-identical file sizes are offset for legibility. Markers show published placement, R is the reference, and orange outlines mark the efficiency frontier.
Score change and file-size difference are measured against the published F16 reference; no quantization claim is inferred.
Bars show absolute percent correct. The orange outline marks the strongest result in each concern for the selected view.
How to read Fluid
Competition results keep their published placement; Fluid-ranked sets keep their Fluid rank. REF is the explicit base control and does not consume a place.
The overall score is supported by seven concern scores, not a single opaque number.
Format validity and length-limit counts remain attached, including completed 0% results.