Sabiá-4: benchmark results
Provider: Other.
Aggregate Rating 1650 ± 40, rank #628 of 10916 rated models, from 11 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Magis-Bench | 5.03 | AVG (self-reported) | 68.2 |
| CAPITU - Easy | 98.6 | Strict prompt-level accuracy (%) on the 70 easy prompts (one | 61.8 |
| CAPITU | 94.4 | Final score (%): the mean over prompts of (2 x the share of | 58.8 |
| CAPITU - Instruction-Level Accuracy | 91.6 | Instruction-level accuracy (%): share of all individual cons | 58.8 |
| CAPITU - Multi-Turn | 78 | Conversation accuracy (%): share of the 100 three-turn CAPIT | 58.8 |
| P3B3 | 87.2 | LLM pt-PT (↑) (self-reported) | 57.9 |
| CAPITU - Strict Accuracy | 84.5 | Strict prompt-level accuracy (%): share of prompts whose con | 52.9 |
| Prosa | 70.2 | Score (self-reported) | 46.7 |
| CAPITU - Medium | 85.7 | Strict prompt-level accuracy (%) on the 70 medium prompts (t | 44.1 |
| CAPITU - Hard | 66.7 | Strict prompt-level accuracy (%) on the 60 hard prompts (thr | 41.2 |
| CAPITU - Coherence | 96 | Coherence score (%): a GPT-4o-mini judge's 0-1 rating of the | 29.4 |
Interactive version: theaggregate.ai/model?slug=sabi-4 · How It Works · Data refreshed daily, snapshot 2026-11-08.