Key Specifications

Vendoropenai
Versiono1-preview
Release Date2024-09-12
Context Window128000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://platform.openai.com/docs/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU85.3%2024-09-125-shotview
HUMANEVAL90.1pass@12024-09-12view
GSM8K92.5%2024-09-120-shot CoTview
MATH71.5%2024-09-120-shot CoTview
BBH84.5%2024-09-123-shot CoTview
GPQA57.7%2024-09-120-shotview
IFEVAL84.4%2024-09-12prompt_strictview
ARC96%2024-09-12challengeview
MUSR64.1%2024-09-120-shotview
WINOGRANDE85.2%2024-09-120-shotview

Pricing

TierPriceCurrency
Input$15 / MtokUSD
Output$60 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://openai.com/api/pricing/ · as of 2024-09-12

Compliance

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

o1 Preview

模型概述

OpenAI o1 预览版推理模型, 128K 上下文, 通过链式思维推理在数学/编码/科学任务上达到博士级表现。

核心规格

厂商版本发布日期上下文窗口输入模态输出模态许可证
Openaio1-preview2024-09-12128KtexttextProprietary

基准测试表现

基准得分单位备注
MMLU (Massive Multitask Language Understanding)85.3%5-shot
HumanEval90.1pass@1
GSM8K (Grade School Math 8K)92.5%0-shot CoT
MATH71.5%0-shot CoT
BBH (BIG-Bench Hard)84.5%3-shot CoT
GPQA57.7%0-shot
IFEval84.4%prompt_strict
ARC96.0%challenge
MUSR64.1%0-shot
WinoGrande85.2%0-shot

定价

输入输出缓存读取缓存写入

每百万 token

优势

  • MMLU 得分 85.3,知识推理能力强。
  • HumanEval 90.1,代码生成能力出色。
  • GSM8K 92.5,数学推理稳健。

劣势

  • 闭源专有模型,不支持自托管。

适用场景

  • 代码生成与调试
  • Agent 工作流与工具调用

参考文献