DataLearner logo
LL

Llama 4 Maverick

Multimodal modelLlama 4

Llama-4-Maverick-17B-128E

Release date: 2025-04-05Updated: 2025-04-15Views: 1,208
Parameters
400B
Context length
1000K
Multilingual
Supported
Reasoning ability
2/5

Llama-4-Maverick-17B-128E is an AI model published by Facebook AI Research Lab, released on 2025-04-05, for Multimodal model, with 400B parameters, and 1000K context length, requiring about 218GB storage, with a 85.50 score on MMLU.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

Llama 4 Maverick

Model basics

Reasoning traces
No data
Thinking modes
Thinking modes not supported
Context length
1000K tokens
Max output length
4K tokens
Model type
Multimodal model
Modality (in / out)
Text, Image, Audio, Video → Text
Release date
2025-04-05
Model file size
218GB
MoE architecture
No
Total params / Active params
400B / Not applicable
Knowledge cutoff
No data
Llama 4 Maverick

Open source & experience

Code license
Weights license
- Commercial use permitted
Live demo
N/A
Llama 4 Maverick

Official resources

Paper
DataLearnerAI blog
N/A
Llama 4 Maverick

API details

API speed
4/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Standard
TypeConditionInputOutput
Text-$0.150/ 1M tokens$0.600/ 1M tokens

“—” means the modality is not billed in that direction, or the vendor has not published a price for it.

Llama 4 Maverick

Benchmark Results

Llama 4 Maverick currently shows benchmark results led by MBPP (22 / 96, score 77.60), MMLU (40 / 124, score 85.50), MMMU (39 / 74, score 73.40). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking
Tool usage

General Knowledge

6 evaluations
Benchmark / mode
Score
Rank/total
MMLU
Standard Mode
85.50
40 / 124
MMLU Pro
Standard Mode
62.90
118 / 176
HLE
Standard Mode
4.90
486 / 563
HLE
unknown
5.68
465 / 563
ARC-AGI-1
Thinking Mode
4.38
144 / 147
ARC-AGI-2
Thinking Mode
0
133 / 136

Coding and Software Engineer

3 evaluations
Benchmark / mode
Score
Rank/total
MBPP
Standard Mode
77.60
22 / 96
LiveCodeBench
Standard Mode
39.70
187 / 250
SciCode
Standard Mode
31.70
122 / 130

Math and Reasoning

3 evaluations
Benchmark / mode
Score
Rank/total
MATH
Standard Mode
61.20
30 / 42
AIME2025
Standard Mode
19.30
199 / 215
FrontierMath
Standard Mode
0.70
55 / 60

General Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Standard Mode
67.10
343 / 462

Multimodal Understanding

3 evaluations
Benchmark / mode
Score
Rank/total
MMMU
unknown
73.40
39 / 74
MMMU-Pro
Standard Mode
62.10
165 / 227
GDP.pdf
Standard Mode
0.80
115 / 118

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleBench
Standard Mode
27.70
76 / 93

Agent Level Benchmark

4 evaluations
Benchmark / mode
Score
Rank/total
τ²-Bench - Telecom
Standard ModeTools
17.80
253 / 264
Aider-Polyglot
Standard Mode
15.60
54 / 59
Terminal Bench Hard
Standard ModeTools
6.80
194 / 244
τ³-Banking
Standard ModeTools
3.70
159 / 164

Instruction Following

1 evaluations
Benchmark / mode
Score
Rank/total
IF Bench
Standard Mode
43
194 / 282

Claw-style Agent Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
Pinch Bench
Thinking ModeTools
46.10
37 / 38

AI Agent - Tool Usage

1 evaluations
Benchmark / mode
Score
Rank/total
Terminal-Bench 2.1
Standard ModeTools
7.90
177 / 192
Llama 4 Maverick

Publisher

Facebook AI Research Lab
Facebook AI Research Lab
View publisher details
Llama-4-Maverick-17B-128E

Model Overview

Llama-4-Maverick-17B-128E is an AI model published by Facebook AI Research Lab, released on 2025-04-05, for Multimodal model, with 400B parameters, and 1000K context length, requiring about 218GB storage, with a 85.50 score on MMLU.

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code