Llama 3.2 90B Vision: Complete Benchmark Performance Guide

meta 3.2-90b-vision 2024-09-25 Llama 3.2 Community License

In-depth analysis of Llama 3.2 90B Vision performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormeta
Version3.2-90b-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
MMLU76.5%2024-09-255-shotview
HUMANEVAL73.1pass@12024-09-25—view
GSM8K82.6%2024-09-250-shot CoTview
MATH52.5%2024-09-250-shot CoTview
BBH76.1%2024-09-253-shot CoTview
GPQA42.2%2024-09-250-shotview
IFEVAL74.5%2024-09-25prompt_strictview
ARC94.6%2024-09-25challengeview
MUSR53.5%2024-09-250-shotview
WINOGRANDE79.3%2024-09-250-shotview

Pricing

TierPriceCurrency
Input$1.2 / MtokUSD
Output$1.2 / 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 90B Vision

모델 개요

Meta Llama 3.2 90B Vision 多模态开源模型, 128K 上下文, 支持图像理解, 性能接近 GPT-4V。

핵심 사양

공급업체버전출시일컨텍스트 창입력 모달리티출력 모달리티라이선스
Meta3.2-90b-vision2024-09-25128Ktext, imagetextLlama 3.2 Community License

벤치마크 성능

벤치마크점수단위비고
MMLU (Massive Multitask Language Understanding)76.5%5-shot
HumanEval73.1pass@1—
GSM8K (Grade School Math 8K)82.6%0-shot CoT
MATH52.5%0-shot CoT
BBH (BIG-Bench Hard)76.1%3-shot CoT
GPQA42.2%0-shot
IFEval74.5%prompt_strict
ARC94.6%challenge
MUSR53.5%0-shot
WinoGrande79.3%0-shot

가격

입력출력캐시 읽기캐시 쓰기
————

백만 토큰당

강점

  • 支持文本、图像、音频多模态输入。

약점

  • 闭源专有模型，不支持自托管。

사용 사례

  • 代码生成与调试
  • 视觉与图像理解

참고문헌