Llama 4 Maverick
Meta AI🇺🇸 United States
Llama 4's multimodal MoE model with 1M context. Matches GPT-4o on most benchmarks.
Context window1000K tokens
Input / 1M tokens$0.27
Output / 1M tokens$0.85
Version History
Llama-4-Maverick-17B-128Emajor
Llama 4 Maverick MoE model with 1M context and native multimodal understanding. Matches GPT-4o on most benchmarks open-weights.
Benchmark Scores
Full leaderboard →94.4%
DocVQA
69.8%
GPQA
88.4%
HumanEval
88.9%
MATH
91.4%
MMLU
73.5%
MMMU
90.0 tokens_per_sec
Speed (tok/s)
51.2%
SWE-bench Verified