Sabiazinho-4: benchmark results
Provider: Other.
Aggregate Rating 1646 ± 31, rank #646 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CAPITU - Easy | 100 | Strict prompt-level accuracy (%) on the 70 easy prompts (one | 85.3 |
| CAPITU - Multi-Turn | 87 | Conversation accuracy (%): share of the 100 three-turn CAPIT | 73.5 |
| CAPITU - Hard | 73.3 | Strict prompt-level accuracy (%) on the 60 hard prompts (thr | 70.6 |
| CAPITU | 94.6 | Final score (%): the mean over prompts of (2 x the share of | 64.7 |
| CAPITU - Instruction-Level Accuracy | 93.1 | Instruction-level accuracy (%): share of all individual cons | 64.7 |
| CAPITU - Strict Accuracy | 87 | Strict prompt-level accuracy (%): share of prompts whose con | 64.7 |
| MARCA (Portuguese, Basic) | 70.5 | Checklist accuracy (0-1, times 100): share of a question's c | 53.8 |
| BLUEX v2 (Official) | 8.58 | Mean rubric score (0-10; 880 subquestions; evaluation errors | 50 |
| MARCA (English, Basic) | 70.9 | Checklist accuracy (0-1, times 100): share of a question's c | 46.2 |
| MARCA (Portuguese, Orchestrator) | 75.3 | Checklist accuracy (0-1, times 100): share of a question's c | 46.2 |
| CAPITU - Medium | 85.7 | Strict prompt-level accuracy (%) on the 70 medium prompts (t | 44.1 |
| Magis-Bench | 4.06 | AVG (self-reported) | 40.9 |
Interactive version: theaggregate.ai/model?slug=sabiazinho-4 · How It Works · Data refreshed daily, snapshot 2026-11-08.