Llama 4 Scout Base: benchmark results
Base Llama 4 Scout checkpoint. Provider: Meta. Released 2025-04-05. Access: Open.
Aggregate Rating 1522 ± 14, rank #707 of 1656 rated models, from 691 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MEGA-Bench Task - Flowchart Code Generation | 77.8 | Task Score (%) | 100 |
| MEGA-Bench Task - OCR Resume Employer Plain | 78.6 | Task Score (%) | 98.8 |
| MEGA-Bench Task - Places365 Scene Type Classification | 100 | Task Score (%) | 98.8 |
| Swallow - Pre-trained English - HellaSwag | 68.9 | Accuracy (%) | 98 |
| Swallow - Pre-trained English - OpenBookQA | 43.2 | Accuracy (%) | 98 |
| Swallow - Pre-trained Japanese - WMT20 Ja-En | 25.8 | BLEU | 98 |
| Open PL LLM - PolQA Open Book (generative, 5-shot) | 91.92 | Levenshtein Similarity (%) | 96.7 |
| EuroEval Swedish NLU - Swerec | 80.1 | Matthews correlation coefficient (%) | 95.9 |
| Open PL LLM - 8TAGS (multiple choice, 5-shot) | 79.85 | Accuracy (%) | 95.6 |
| MEGA-Bench Task - Annoying Word Search | 0.4 | Task Score (%) | 95.3 |
| Open PL LLM - PoQuAD Open Book (generative, 5-shot) | 72.38 | Levenshtein Similarity (%) | 95.1 |
| MEGA-Bench Task - Camera Trajectory To Video Selection | 64.3 | Task Score (%) | 94.2 |
Interactive version: theaggregate.ai/model?slug=llama-4-scout-base · How It Works · Data refreshed daily, snapshot 2026-10-20.