Llama 4 Maverick: benchmark results
Meta's open Llama 4 Maverick, a 400B sparse MoE (17B active) multimodal model (April 2025). Provider: Meta. Released 2025-04-05. Access: Open.
Aggregate Rating 1593 ± 3, rank #464 of 1656 rated models, from 1336 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| 6G-Bench | 82.9 | Accuracy (%) on the full set of 6G-Bench's 3,722 expert-vali | 100 |
| 6G-Bench - AI-Native Networking and Agentic Control | 85.5 | Accuracy (%) on the AI-native networking and agentic control | 100 |
| 6G-Bench - Distributed Intelligence and Emerging Use Cases | 80.6 | Accuracy (%) on the distributed intelligence and emerging 6G | 100 |
| 6G-Bench - Network Slicing and Resource Management | 80.5 | Accuracy (%) on the network slicing and resource management | 100 |
| AMVICC - MMVP Individual Accuracy - Quantity and Count | 95.83 | Individual accuracy (%; the 24 MMVP questions on 12 image pa | 100 |
| AMVICC - MMVP Individual Accuracy - State and Condition | 88.89 | Individual accuracy (%; the 54 MMVP questions on 27 image pa | 100 |
| AMVICC - MMVP Pair Accuracy - Quantity and Count | 91.67 | Pair accuracy (%; the 12 MMVP image pairs in the Quantity an | 100 |
| AuditBench - Lab Attack Classification | 100 | F1 (0-100) for attack classification (is a log chunk malicio | 100 |
| AuditBench - Lab Lateral Movement | 62 | F1 (0-100) for lateral-movement identification on the Lab da | 100 |
| Confidence Decision Faithfulness - GSM8K - Policy Consistency | 1 | Policy Consistency (0-1) | 100 |
| MedRepBench - Abnormal Flag Recall | 75.83 | Field-level recall (%; abnormality flag of each ground-truth | 100 |
| MedRepBench - Abnormal Flag Recall (OCR-Assisted) | 86.56 | Field-level recall (%; abnormality flag of each ground-truth | 100 |
Interactive version: theaggregate.ai/model?slug=llama-4-maverick · How It Works · Data refreshed daily, snapshot 2026-10-20.