Llama 3.2 11B Vision: Complete Benchmark Performance Guide

meta 3.2-11b-vision 2024-09-25 Llama 3.2 Community License

In-depth analysis of Llama 3.2 11B Vision performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormeta
Version3.2-11b-vision
Release Date2024-09-25
Context Window128000 tokens
Input Modalitiestext, image
Output Modalitiestext
LicenseLlama 3.2 Community License
Documentationhttps://llama.meta.com/docs/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU61.9%2024-09-255-shotview
HUMANEVAL50.1pass@12024-09-25—view
GSM8K68.2%2024-09-250-shot CoTview
MATH29.9%2024-09-250-shot CoTview
BBH69.1%2024-09-253-shot CoTview
GPQA35.5%2024-09-250-shotview
IFEVAL58.4%2024-09-25prompt_strictview
ARC86%2024-09-25challengeview
MUSR41.7%2024-09-250-shotview
WINOGRANDE72.7%2024-09-250-shotview

Pricing

TierPriceCurrency
Input$0.55 / MtokUSD
Output$0.55 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://ai.meta.com/blog/ · as of 2024-09-25

Compliance

  • Data Residency: self-host
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

Llama 3.2 11B Vision

Model Overview

Meta Llama 3.2 11B Vision 经济型多模态开源模型, 128K 上下文, 11B 参数, 适合边缘视觉任务。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Meta3.2-11b-vision2024-09-25128Ktext, imagetextLlama 3.2 Community License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)61.9%5-shot
HumanEval50.1pass@1—
GSM8K (Grade School Math 8K)68.2%0-shot CoT
MATH29.9%0-shot CoT
BBH (BIG-Bench Hard)69.1%3-shot CoT
GPQA35.5%0-shot
IFEval58.4%prompt_strict
ARC86.0%challenge
MUSR41.7%0-shot
WinoGrande72.7%0-shot

Pricing

InputOutputCache ReadCache Write
————

per million tokens

Strengths

  • Input Modalities: text, image, audio.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Vision and image understanding

References