Llama 3.2 3B: Complete Benchmark Performance Guide

meta 3.2-3b 2024-09-25 Llama 3.2 Community License

In-depth analysis of Llama 3.2 3B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormeta
Version3.2-3b
Release Date2024-09-25
Context Window128000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseLlama 3.2 Community License
Documentationhttps://llama.meta.com/docs/

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 52.5 % 2024-09-25 5-shot view
HUMANEVAL 42.5 pass@1 2024-09-25 — view
GSM8K 46.4 % 2024-09-25 0-shot CoT view
MATH 10 % 2024-09-25 0-shot CoT view
BBH 46 % 2024-09-25 3-shot CoT view
GPQA 18.8 % 2024-09-25 0-shot view
IFEVAL 48.6 % 2024-09-25 prompt_strict view
ARC 76.3 % 2024-09-25 challenge view
MUSR 29.3 % 2024-09-25 0-shot view
WINOGRANDE 67.8 % 2024-09-25 0-shot view

Pricing

Tier Price Currency
Input$0.15 / MtokUSD
Output$0.15 / 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 3B

Model Overview

Meta Llama 3.2 3B 轻量开源模型, 128K 上下文, 3B 参数, 适合移动设备与边缘部署。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Meta 3.2-3b 2024-09-25 128K text text Llama 3.2 Community License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 52.5 % 5-shot
HumanEval 42.5 pass@1 —
GSM8K (Grade School Math 8K) 46.4 % 0-shot CoT
MATH 10.0 % 0-shot CoT
BBH (BIG-Bench Hard) 46.0 % 3-shot CoT
GPQA 18.8 % 0-shot
IFEval 48.6 % prompt_strict
ARC 76.3 % challenge
MUSR 29.3 % 0-shot
WinoGrande 67.8 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • MMLU 52.5, weak knowledge reasoning.
  • HumanEval 42.5, coding weak.
  • Proprietary, not self-hostable.

Use Cases

  • General chat and Q&A

References