Llama 3.2 11B Instruct: benchmark results
Meta Llama 3.2 11B vision instruction-tuned checkpoint. Provider: Meta. Released 2024-09-25. Access: Open.
Aggregate Rating 1424 ± 6, rank #1126 of 1656 rated models, from 981 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| KazakhOCR (Latin) | 21.6 | Character error rate of the transcription of Latin-script Ka | 100 |
| ThermEval - Human Counting (FLIR) | 2.84 | Mean absolute error (people, 0 or more) on Task 3, the numbe | 100 |
| AGC-Bench - schnovel | 1.91 | Dataset z-score | 98.8 |
| Open LMM Reasoning - WeMath - InsufficientKnowledge (Loose) | 80 | Accuracy (%) | 97.5 |
| Open LMM Reasoning - WeMath - RoteMemorization (Loose) | 36 | Accuracy (%) | 97.5 |
| ThermEval - Modality Identification (FLIR) | 100 | Accuracy (times 100, 0-100) on Task 1, thermal or RGB, on ba | 97.1 |
| OpenVLM MMT-Bench - Abstract Visual Recognition | 90 | Score (%) | 96.8 |
| AGC-Bench - puntuguese | 1.02 | Dataset z-score | 96.3 |
| ThermEval - Thermal Reasoning (Two People) | 61 | Accuracy (times 100, 0-100) on Task 5, whether the chest, fo | 95.7 |
| OpenVLM MMT-Bench - National Flag Recognition | 100 | Score (%) | 94.9 |
| OpenVLM MMT-Bench - Table Structure Recognition | 100 | Score (%) | 94.7 |
| OpenVLM MMT-Bench - Logo and Brand Recognition | 100 | Score (%) | 94.4 |
Interactive version: theaggregate.ai/model?slug=llama-3-2-11b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.