Llama 2 7B: Complete Benchmark Performance Guide

meta 2-7b 2023-07-18 Llama 2 Community License

In-depth analysis of Llama 2 7B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormeta
Version2-7b
Release Date2023-07-18
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseLlama 2 Community License
Documentationhttps://llama.meta.com/docs/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU70.8%2023-07-185-shotview
HUMANEVAL42.6pass@12023-07-18—view
GSM8K31%2023-07-180-shot CoTview
MATH19.4%2023-07-180-shot CoTview
BBH54.7%2023-07-183-shot CoTview
GPQA23.1%2023-07-180-shotview
IFEVAL58.5%2023-07-18prompt_strictview
ARC86.5%2023-07-18challengeview
MUSR35.3%2023-07-180-shotview
WINOGRANDE77.2%2023-07-180-shotview

Pricing

TierPriceCurrency
Input$0.2 / MtokUSD
Output$0.2 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://ai.meta.com/blog/ · as of 2023-07-18

Compliance

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

Llama 2 7B

Model Overview

Meta Llama 2 7B 经济型开源模型, 4K 上下文, 7B 参数, 广泛用于本地部署与微调基座。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Meta2-7b2023-07-184KtexttextLlama 2 Community License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)70.8%5-shot
HumanEval42.6pass@1—
GSM8K (Grade School Math 8K)31.0%0-shot CoT
MATH19.4%0-shot CoT
BBH (BIG-Bench Hard)54.7%3-shot CoT
GPQA23.1%0-shot
IFEval58.5%prompt_strict
ARC86.5%challenge
MUSR35.3%0-shot
WinoGrande77.2%0-shot

Pricing

InputOutputCache ReadCache Write
————

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • HumanEval 42.6, coding weak.
  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • General chat and Q&A

References