Meta

Llama 4 Maverick

Multimodal
Zero-eval
#1MGSM
#1TydiQA
#2ChartQA

by Meta

+
+
+
+
About

Llama 4 Maverick was developed as a variant in Meta's fourth-generation language model family, designed to explore specialized capabilities and training approaches. Built to push the boundaries of open-source model development, it represents experimentation with advanced techniques in the Llama lineage.

+
+
+
+
Pricing Range
Input (per 1M)$0.17 -$0.63
Output (per 1M)$0.60 -$1.79
Providers7
+
+
+
+
Timeline
AnnouncedApr 5, 2025
ReleasedApr 5, 2025
+
+
+
+
Specifications
Training Tokens22.0T
Capabilities
Multimodal
+
+
+
+
License & Family
License
Llama 4 Community License Agreement
Performance Overview
Performance metrics and category breakdown

Overall Performance

13 benchmarks
Average Score
71.8%
Best Score
94.4%
High Performers (80%+)
5

Performance Metrics

Max Context Window
2.0M
Avg Throughput
193.4 tok/s
Avg Latency
1ms
+
+
+
+
All Benchmark Results for Llama 4 Maverick
Complete list of benchmark scores with detailed information
DocVQA
multimodal
0.94
94.4%
Self-reported
MGSM
text
0.92
92.3%
Self-reported
ChartQA
multimodal
0.90
90.0%
Self-reported
MMLU
text
0.85
85.5%
Self-reported
MMLU-Pro
text
0.81
80.5%
Self-reported
MBPP
text
0.78
77.6%
Self-reported
MathVista
multimodal
0.74
73.7%
Self-reported
MMMU
multimodal
0.73
73.4%
Self-reported
GPQA
text
0.70
69.8%
Self-reported
MATH
text
0.61
61.2%
Self-reported
Showing 1 to 10 of 13 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+