Meta/Llama 4/Apr 5, 2025
Llama 4 Maverick
Meta's flagship open-weight model. A 17B active parameter model with 128 MoE experts. Beats GPT-4o and Gemini 2.0 Flash across a broad range of benchmarks. Scores 73.4% on MMMU and 73.7% on MathVista. Fully open-weight and available through multiple API providers.
textcodereasoningvision
Arena ELO
1,327
Input Price
$0.31/M
Output Price
$0.85/M
Speed
126 t/s
Context
1M
Latency
810ms
Capability Assessment
GPQA Diamond69.8%
MMMU Pro73.4%
Comparative Analysis
| Metric | Llama 4 Maverick | Claude Opus 4.6 | Gemini 3.1 Pro | Grok 4 |
|---|---|---|---|---|
| SWE-bench | — | 80.8% | 80.6% | 72.0% |
| AIME 2025 | — | 100.0% | 91.2% | 94.0% |
| GPQA Diamond | 69.8% | 91.3% | 94.3% | 88.0% |
| MMLU | 92.4% | 91.0% | 92.6% | — |
| Input $/M | $0.31 | $5 | $2 | $3 |
| Output $/M | $0.85 | $25 | $12 | $15 |