Llama 4 Maverick
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.
Input
Output
Context
1049K
Max Output
16K
Parameters
400B
Technical Specifications
Model TypeVision
Context Window1.0M
Max Output Tokens16K
Parameters400B
Training Cutoff—
Licensellama-3.1
Capabilities
Input Modalities
textimage
Output Modalities
text
Features
code_generationfunction_callingjson_modereasoningstreamingtext_generationvision
Benchmarks
4.8%
HLE
39.0%
AIME
67.1%
GPQA
33.1%
SciCode
88.9%
MATH 500
80.9%
MMLU Pro
19.3
Math
15.6
Coding
39.7%
LiveCodeBench
18.4
Intelligence
110
Speed (tok/s)
0.607
TTFA (s)
0.607
TTFT (s)
Resources & Links
Estimated GPU Requirements for Llama 4 Maverick
Estimates based on INT8 quantization. Actual requirements vary by framework and configuration.
Related Tools
Data sourced from official provider APIs and documentation
Last updated: May 5, 2026
Start building with the right model.
From model selection to production, one platform, no fragmentation.