01Overview02Models03Rankings04Compare05Quiz06Guide

Model

Llama 4 Maverick NO.22

Open natively-multimodal MoE flagship — a strong, cheap general workhorse.

CompareBack to the index

Meta AI · Llama · Open · Open weights

Specification

Fig.01 — Llama — Meta AI

USA

Context window
1.0Mtokens
Max output
16Ktokens
Input price
$0.17/ 1M
Output price
$0.60/ 1M
Throughput
—
Class
Open
Modalities
Text, Image
Weights
Open
Parameters
400B
License
Llama 4 Community License
Knowledge cutoff
Aug 2024
Released
Apr 2025

Overview

Llama 4 Maverick is Meta's open-weight flagship, a 400B-total / 17B-active mixture-of-experts model with native text+image understanding and a 1M-token context. It is an excellent, cheap, self-hostable general assistant and a favorite base for fine-tuning, but it trails the closed frontier on long-horizon agentic coding. Best treated as a high-value open workhorse rather than an autonomous coding agent.

Strengths

  • Open weights, self-hostable and fine-tune-friendly
  • Native multimodal (text + image) at low cost
  • Very large 1M-token context
  • Strong multilingual coverage

Trade-offs

  • Unreliable on long autonomous agentic coding runs
  • Behind the frontier on hard reasoning/coding
  • Community license restricts 700M+ MAU use
  • Vision weaker than Gemini/GPT flagships

Fit

Best for

  • Self-hosted general assistants
  • Fine-tuning and domain adaptation
  • High-volume multilingual chat
  • Cost-sensitive multimodal apps

Not ideal for

  • Long autonomous coding agents
  • Frontier math/reasoning tasks

Capabilities

Ten axes, normalized 0–100 and scored the same way across the whole catalog — so a 78 here means what a 78 means anywhere else on the bench.

Fig.02 — Capability radar

0 — 100

Reasoning
68
Coding
62
Math
64
Writing
72
Knowledge
78
Speed
82
Agentic
58
Vision
72
Multilingual
80
Long Context
82

Benchmarks

Public results, with independent third-party runs marked. Bars normalize percentages against 100 and Elo ratings against a 1500 ceiling.

Independent index

Artificial Analysis Intelligence Index

Composite of ~9–10 independent evals · Artificial Analysis

14/ 100

BenchmarkResult
Artificial Analysis Intelligence IndexIndependentGeneral · Artificial Analysis
14%
τ-benchIndependentAgentic · τ-bench
68.5%
MATHMath · Vendor-reported
89.4%
MMLU-ProKnowledge · Vendor-reported
80.5%
GPQA DiamondIndependentReasoning · Independent aggregators
69.8%
MMMUVision · Vendor-reported
73.4%
LiveCodeBenchIndependentCoding · LiveCodeBench
43.4%
LMArena EloIndependentGeneral · LMArena
1417Elo

Alternatives

Models in roughly the same class — the ones worth weighing against this record.

Compare side by side
01Mistral Large 3Mistral AIEurope's flagship open-weight MoE — a strong, cheap, multilingual generalist.02DeepSeek-V3.2DeepSeekUltra-cheap open MoE with standout math and efficiency.03Qwen3.7-MaxAlibaba QwenAlibaba's proprietary agentic reasoning flagship with a 1M context.

Index

PreviousGrok 4.1 FastNextLlama 4 Scout
All models/EDUCATION