Meta AI · USA · Released 2025-04

Llama 4 Maverick

Meta's flagship MoE model — 400B total / 17B active — with native multimodal input and 1M-token context.

Official site GitHub

Access

open-weights

Context

1,000,000 tokens

Modalities

text, vision

Languages

12 (official)

Benchmark scores

MMLU

85.5%

Massive Multitask Language Understanding

GPQA

69.8%

Graduate-level science QA (Diamond)

HumanEval

88.0%

Python code completion

SWE-bench

N/A

Real GitHub issue resolution (Verified)

LiveCodeBench

N/A

Contamination-free coding

AIME

N/A

American Invitational Mathematics Exam

MMMU

73.4%

Multimodal college-level reasoning

MATH

78.0%

Competition-level math word problems

Scores as reported by the vendor or leading public leaderboards. "N/A" means the score has not been publicly disclosed for this metric.

Explain like I'm 5

A giant open model that sees images and reads huge documents.

Key features

Mixture-of-Experts
Native multimodal
1M context
Open weights

Strengths

Open weights
Great cost/perf
Long context

Limitations

Below closed frontier on hardest reasoning

Best for

Self-hosted assistantsFine-tuningOn-prem RAG

Related models

OpenAI

GPT-5

OpenAI's flagship reasoning + multimodal model with a unified router that switches between fast and deep-thinking modes.

Anthropic

Claude Sonnet 4.5

Anthropic's best coding + agentic model, purpose-built for long-horizon computer-use and software engineering tasks.

Google DeepMind

Gemini 2.5 Pro

Google's most capable model with native 1M-token context and full multimodal I/O across text, images, audio and video.

xAI

Grok 4

xAI's frontier reasoning model with real-time X data grounding and a heavy multi-agent variant (Grok 4 Heavy).