LIVE COMPONENT FAMILY
Data specimens
Benchmark cards, tables, chips, and bar summaries keep the report pages dense without turning them into dashboard furniture.
Benchmark cards
A published benchmark is one complete card link. A pending benchmark is text, not an inactive pseudo-button.
Sortable result table
| Model | |
|---|---|
| Fable 5.1 | 38.8% |
| GPT-6 Astra | 33.8% |
| Gemini 3.8 Flash | 31.2% |
Click the header to reorder rows and update aria-sort. Real-SWE uses the same local fixture behavior.
Metadata chips and bar rows
15 models4 languageshuman-verified ground truth
Chips label a bounded set. Bar values remain available as plain text; do not rely on bar length alone.
Language result chart
Spanish, Mexican Spanish family call
These values are drawn from the Speech-to-Text report. Bars are supplemental; labels and numbers remain in the DOM.
Language select
Showing Spanish results.
Select a language to update the chart title, values, and status. This local control adds a testable composition around observed report data.