Sobre

Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward pass (400B total). It supports multilingual text and image input, and produces multilingual text and code output across 12 supported languages. Optimized for vision-language tasks, Maverick is instruction-tuned for assistant-like behavior, image reasoning, and general-purpose multimodal interaction. Maverick features early fusion for native multimodality and a 1 million token context window. It was trained on a curated mixture of public, licensed, and Meta-platform data, covering ~22 trillion tokens, with a knowledge cutoff in August 2024. Released on April 5, 2025 under the Llama 4 Community License, Maverick is suited for research and commercial applications requiring advanced multimodal understanding and high model throughput.

Família do Modelo

Meta: Llama Guard 4 12B2025-04-30 Llama 4 Scout2025-04-05 Llama 4 Scout 17B 16E Instruct2025-04-02 Llama Guard 3 8B2025-02-12 Llama 3.3 Instruct 70B2024-12-06 Meta: Llama 3.3 70B Instruct2024-12-06 Meta: Llama 3.3 70B Instruct (free)2024-12-06 Llama 3.2 Instruct 90B (Vision)2024-09-25

Benchmarks

MMLU-Pro

80.9%

GPQA Diamond

67.1%

HLE

4.8%

LiveCodeBench

39.7%

SciCode

33.1%

TerminalBench Hard

6.8%

MATH-500

88.9%

AIME

39.0%

AIME 2025

19.3%

IFBench

43.0%

Long Context Recall

46.0%

Tau2

17.8%

Média do MercadoMelhor Score

Llama 4 Maverick

Sobre

Família do Modelo

Posição no Mercado

Preços

Calculadora de Custo

vs. Modelos Similares

Desempenho

Benchmarks

Open Source

Comparação Rápida

Modelos Similares