Llama 3.2 Instruct 90B (Vision)

Llama 3.2 Instruct 90B (Vision)

Meta · Released Sep 25, 2024 Professional
Intelligence #14 / 590
82.0 Our Score
Speed #193 / 279
60.0 tokens / sec
Input #481 / 592
$1.38 per 1M tokens
Output #359 / 592
$1.38 per 1M tokens
Context
— Not reported

Analysis Summary

Llama 3.2 Instruct 90B (Vision) is Meta's large vision-language model from late 2024, offering image understanding alongside text generation at a relatively low price of $1.38 per million tokens for both input and output. Its MMLU Pro score of 67.1% and GPQA of 43.2% reflect competent but not exceptional general reasoning, and its livecodebench score of 21.4% limits its usefulness for serious coding tasks.

The vision capability is a genuine differentiator for businesses needing image analysis, document parsing, or multimodal content workflows at moderate cost. However, with an intelligence index of just 6.2, it falls far short of current mid-tier models for complex reasoning, agentic tasks, or high-stakes content generation. It is best suited to lighter workloads where vision is required and budget is a constraint.

For teams that need a cost-effective multimodal model for straightforward tasks, this remains a viable option. Teams with more demanding requirements should look to more recent models with stronger reasoning benchmarks.

Assessed June 30, 2026

Editorial notes

Llama 3.2 90B Vision from Meta brings multimodal capability and a competitive price point, but its intelligence index of 6.2 places it well below current frontier models for demanding business tasks.

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?

Performance Profile

Intelligence1.8Technical0Value6Content0
Intelligence 1.8/10
Technical 0/10
Content 0/10
Value 6/10

How Llama 3.2 Instruct 90B (Vision) compares

Llama 3.2 Instruct 90B (Vision) ranks #306 of 385 AI models we track for overall intelligence. At $1.38 per million input tokens it is cheaper than 19% of comparable models.

Performance Indices

Source: Artificial Analysis

6.2 Intelligence Index

Benchmark Scores

Intelligence

GPQA Diamond 43.2% Graduate-level scientific reasoning
HLE 4.9% Humanity's Last Exam
MMLU Pro 67.1% Multi-task language understanding
MATH 500 62.9% Mathematical problem-solving
AIME 5% Competition mathematics
SciCode 24% Scientific computing

Technical

LiveCodeBench 21.4% Live coding evaluation

Benchmark data from Artificial Analysis and Hugging Face

How does Llama 3.2 Instruct 90B (Vision) stack up?

Compare side-by-side with other professional models.

Compare Models

Model Information

ProviderMeta
Release Date September 25, 2024
Status Active

Pricing

Token Type Cost per 1M tokens Cost per 1K tokens
Input $1.38 $0.001380
Output $1.38 $0.001380

Leaderboard Categories

Frequently asked questions about Llama 3.2 Instruct 90B (Vision)

How much does Llama 3.2 Instruct 90B (Vision) cost?

Llama 3.2 Instruct 90B (Vision) costs $1.38 per million input tokens and $1.38 per million output tokens.

Who created Llama 3.2 Instruct 90B (Vision)?

Llama 3.2 Instruct 90B (Vision) is developed by Meta and was released on September 25, 2024.