Llama 4 Scout
Multimodal
Zero-eval
#2TydiQA
by Meta
+
+
+
+
About
Llama 4 Scout was created as an exploratory variant in the Llama 4 family, designed to investigate new architectures and optimization strategies. Built as part of Meta's commitment to advancing open-source AI, it serves as a testbed for innovations that may inform future model releases.
+
+
+
+
Pricing Range
Input (per 1M)$0.08 -$0.18
Output (per 1M)$0.30 -$0.60
Providers6
+
+
+
+
Timeline
AnnouncedApr 5, 2025
ReleasedApr 5, 2025
+
+
+
+
Specifications
Training Tokens40.0T
Capabilities
Multimodal
+
+
+
+
License & Family
License
Llama 4 Community License Agreement
Performance Overview
Performance metrics and category breakdown
Overall Performance
12 benchmarks
Average Score
67.3%
Best Score
94.4%
High Performers (80%+)
3Performance Metrics
Max Context Window
20.0MAvg Throughput
214.1 tok/sAvg Latency
1ms+
+
+
+
All Benchmark Results for Llama 4 Scout
Complete list of benchmark scores with detailed information
| DocVQA | multimodal | 0.94 | 94.4% | Self-reported | |
| MGSM | text | 0.91 | 90.6% | Self-reported | |
| ChartQA | multimodal | 0.89 | 88.8% | Self-reported | |
| MMLU | text | 0.80 | 79.6% | Self-reported | |
| MMLU-Pro | text | 0.74 | 74.3% | Self-reported | |
| MathVista | multimodal | 0.71 | 70.7% | Self-reported | |
| MMMU | multimodal | 0.69 | 69.4% | Self-reported | |
| MBPP | text | 0.68 | 67.8% | Self-reported | |
| GPQA | text | 0.57 | 57.2% | Self-reported | |
| MATH | text | 0.50 | 50.3% | Self-reported |
Showing 1 to 10 of 12 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+