Llama 3 70B: Complete Benchmark Performance Guide

meta 3-70b 2024-04-18 Llama 3 Community License

In-depth analysis of Llama 3 70B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormeta
Version3-70b
Release Date2024-04-18
Context Window8192 tokens
Input Modalitiestext
Output Modalitiestext
LicenseLlama 3 Community License
Documentationhttps://llama.meta.com/docs/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU79.5%2024-04-185-shotview
HUMANEVAL73pass@12024-04-18—view
GSM8K76.2%2024-04-180-shot CoTview
MATH50.1%2024-04-180-shot CoTview
BBH77.6%2024-04-183-shot CoTview
GPQA38.2%2024-04-180-shotview
IFEVAL72.2%2024-04-18prompt_strictview
ARC93.4%2024-04-18challengeview
MUSR51.2%2024-04-180-shotview
WINOGRANDE79.2%2024-04-180-shotview

Pricing

TierPriceCurrency
Input$0.9 / MtokUSD
Output$0.9 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://ai.meta.com/blog/ · as of 2024-04-18

Compliance

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

Llama 3 70B

Model Overview

Meta Llama 3 70B 中端开源模型, 8K 上下文, 在 MMLU/HumanEval 上接近 GPT-4, 开源旗舰之一。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Meta3-70b2024-04-188KtexttextLlama 3 Community License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)79.5%5-shot
HumanEval73.0pass@1—
GSM8K (Grade School Math 8K)76.2%0-shot CoT
MATH50.1%0-shot CoT
BBH (BIG-Bench Hard)77.6%3-shot CoT
GPQA38.2%0-shot
IFEval72.2%prompt_strict
ARC93.4%challenge
MUSR51.2%0-shot
WinoGrande79.2%0-shot

Pricing

InputOutputCache ReadCache Write
————

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.
  • Context window 8K is limited.

Use Cases

  • Code generation and debugging

References