GPT-4 32K: Complete Benchmark Performance Guide

openai 4-32k 2023-03-14 Proprietary

In-depth analysis of GPT-4 32K performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoropenai
Version4-32k
Release Date2023-03-14
Context Window32768 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://platform.openai.com/docs/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU80.8%2023-03-145-shotview
HUMANEVAL76.8pass@12023-03-14—view
GSM8K81.6%2023-03-140-shot CoTview
MATH40.1%2023-03-140-shot CoTview
BBH79.2%2023-03-143-shot CoTview
GPQA40.9%2023-03-140-shotview
IFEVAL74.7%2023-03-14prompt_strictview
ARC95%2023-03-14challengeview
MUSR54.9%2023-03-140-shotview
WINOGRANDE82.2%2023-03-140-shotview

Pricing

TierPriceCurrency
Input$60 / MtokUSD
Output$120 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://openai.com/api/pricing/ · as of 2023-03-14

Compliance

  • Data Residency: US
  • SOC2: ✓
  • HIPAA: ✓
  • GDPR: ✓
  • ISO 27001: ✓

GPT-4 32K

Aperçu du modèle

OpenAI GPT-4 32K 上下文版本, 支持 32768 token 长文本输入, 适合长文档理解与生成任务。

Spécifications principales

FournisseurVersionDate de sortieFenêtre de contexteModalités d’entréeModalités de sortieLicence
Openai4-32k2023-03-1432KtexttextProprietary

Performance aux benchmarks

BenchmarkScoreUnitéNotes
MMLU (Massive Multitask Language Understanding)80.8%5-shot
HumanEval76.8pass@1—
GSM8K (Grade School Math 8K)81.6%0-shot CoT
MATH40.1%0-shot CoT
BBH (BIG-Bench Hard)79.2%3-shot CoT
GPQA40.9%0-shot
IFEval74.7%prompt_strict
ARC95.0%challenge
MUSR54.9%0-shot
WinoGrande82.2%0-shot

Tarification

EntréeSortieLecture cacheÉcriture cache
————

par million de jetons

Forces

  • MMLU score 80.8, strong knowledge reasoning.
  • 上下文窗口 32K，支持长文本。

Faiblesses

  • 闭源专有模型，不支持自托管。

Cas d’usage

  • 代码生成与调试
  • 长文档摘要
  • Agent 工作流与工具调用

Références