Meta

Llama 4 Maverick

by Metatext+image->text5 endpoints
Neura Intelligence Index
51.2/ 100
Rank
#173
Confidence
low
Benchmarks
3(8% coverage)

Overview

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...

Capabilities

Text generation
Image understanding
Tool use / Function calling
Structured output
Long context
Top-P sampling
Stop sequences
Deterministic seed
Log probabilities

Modalities

Input
TextImage
Output
Text

Technical Specifications

Context Window
1.0M tokens
Max Output
16.4K tokens
Knowledge Cutoff
2024-08-31
Tokenizer
Llama4
Uptime (24h)
9998.6%
Parameters
400000000000B
License
llama_4_community_license_agreementOpen Weights
Country
🇺🇸US
Release Date
April 5, 20251y

Supported Parameters

frequency_penaltylogit_biaslogprobsmax_tokensmin_ppresence_penaltyrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p

Benchmark Performance

reasoning

BenchmarkScoreSource
GPQA Diamond
69.8
verified

multimodal

BenchmarkScoreSource
MMMU
73.4
verified
MMMU-Pro
59.6
verified

Performance Over Time

Pricing

Per 1M tokens
Input$0.19
Output$0.65
Added
April 5, 2025
Last synced 9/20/2026