Multimodal

LLaMA 3.2 90B Vision Environmental Impact

StandardEstimated

Large open-source multimodal model rivalling GPT-4V

Architecture
Vision-Language Transformer (decoder-only)
Parameters
90B
Context
128,000 tokens
Provider
Meta
1.5 Wh
Energy per query
0.60 g
CO₂ per query
3 mL
Water per query
5x more than
vs Google search

Energy per query

1.5 Wh

5x more than a Google search (0.3 Wh)

CO2 per query

0.60 g

Global Average grid (475 gCO₂/kWh)

Water per query

3 mL

~357 queries to fill 1 litre

Processing location

Self-hosted (varies, requires multi-GPU)

Provider

Meta

Category

Multimodal

Grid carbon intensity

475 g CO2/kWh (27% renewable)

How does LLaMA 3.2 90B Vision compare?

Ranked #75 of 152 models by energy per query

0 Wh0.4 Wh0.8 Wh1.2 Wh1.6 WhPhi-4MultimodalLLaMA 3.2 11BVisionLLaMA 3.2 90BVisionQwen 3.5 OmniGoogle search (0.3 Wh)

Detailed Breakdown

Energy Consumption

LLaMA 3.2 90B Vision is the largest open-source multimodal model, consuming approximately 1.5 Wh per query. It approaches GPT-4V-class capabilities for image understanding while being fully open-weight. Requires multi-GPU setups for inference.

Power Source & Carbon

Requires cloud GPU clusters for inference due to its size. Available through major cloud providers and specialised inference platforms.

Water Usage

At ~2.8 mL per query, the 90B vision model has a moderate water footprint.

About LLaMA 3.2 90B Vision

LLaMA 3.2 90B Vision is a multimodal model from Meta, released in September 25, 2024. Large open-source multimodal model rivalling GPT-4V. Each query uses 1.5 Wh of energy and produces 0.60 g of CO₂. That's 5x the energy of a Google search — reflecting the computational demands of multimodal.

These figures are estimates derived from hardware specifications and API benchmarks — Meta has not published official energy data for LLaMA 3.2 90B Vision. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.

LLaMA 3.2 90B Vision in Context

92%
potential savings

The efficiency alternative

Phi-4 Multimodal performs the same type of task using just 0.12 Wh per query — 92% less energy than LLaMA 3.2 90B Vision. For a user sending 25 queries per day, switching would save 12.6 kWh per year.

Key Insights

Open-source weights — can be self-hosted on infrastructure you control

What does your LLaMA 3.2 90B Vision usage cost the planet?

Use our calculator to estimate your personal environmental footprint based on how often you use LLaMA 3.2 90B Vision.

Calculate My Compute

Frequently Asked Questions

How much energy does LLaMA 3.2 90B Vision use per query?

Each LLaMA 3.2 90B Vision query consumes approximately 1.5 Wh of energy. This is 5x more than a traditional Google search (~0.3 Wh).

What is LLaMA 3.2 90B Vision's carbon footprint?

Based on the carbon intensity of Self-hosted (varies, requires multi-GPU), each query produces approximately 0.60 g of CO2. The grid in this region has a carbon intensity of 475 g CO2/kWh with 27% renewable energy.

How much water does LLaMA 3.2 90B Vision use?

Each query consumes approximately 3 mL of water, primarily used for cooling the data centers that process the request.

How does LLaMA 3.2 90B Vision compare to a Google search?

A LLaMA 3.2 90B Vision query uses 5x more than a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while LLaMA 3.2 90B Vision uses 1.5 Wh.

Technical Details

Architecture

Vision-Language Transformer (decoder-only)

Parameters

90B

Context window

128,000 tokens

Release date

2024-09-25

Open source

Yes

Training data cutoff

2024-08