Llama 4 Maverick

Meta (via hosted providers)

open_source_flagship

Capabilities

VisionToolsCodeLong ContextStructured OutputMultilingual

Strengths

Open weights — full control
1M context
Strong multilingual
Can self-host

Weaknesses

Quality varies by provider
Self-hosting needs GPUs
Meta licence has restrictions

Pricing

Input / 1M tokens

$0.15

Output / 1M tokens

$0.6

Context window

1,000k

Transparency

Open weights

7.0 / 10

Open training data

3.0 / 10

Open methodology

5.0 / 10

Licence openness

7.0 / 10

Provider disclosure

3.0 / 10

FMTI company score

300.0 / 10

Composite:5.0 / 10

Open weights with registration. Meta Community Licence restricts competitors. FMTI 30 — open model but opaque company.

Sustainability

Inference energy

5.5 / 10

Training footprint

N/A

Provider infrastructure

8.0 / 10

Composite:6.8 / 10

Meta claims 100% renewable operations (bucket 0.8). Net-zero by 2030. Water-positive by 2030. Dry cooling at new facilities. MoE 17B active = efficient inference.

Task Fitness

question answering

78%

code

84%

mathematics

81%

business communication

66%

analysis

78%

extraction

80%

long-form generation

78%

research

55%

reasoning

78%

summarisation

82%

translation

85%

conversation

78%