核心规格

厂商cohere
版本embed-english-v3
发布日期2023-11-01
上下文窗口512 tokens
输入模态text
输出模态
许可CC-BY-NC-4.0
文档https://docs.cohere.com/docs

基准测试表现

基准得分单位评测日期备注来源
MMLU0%2023-11-01embed model - not applicable查看
HUMANEVAL0pass@12023-11-01embed model - not applicable查看
GSM8K0%2023-11-01embed model - not applicable查看
MATH0%2023-11-01embed model - not applicable查看
BBH0%2023-11-01embed model - not applicable查看
GPQA0%2023-11-01embed model - not applicable查看
IFEVAL0%2023-11-01embed model - not applicable查看
ARC0%2023-11-01embed model - not applicable查看
MUSR0%2023-11-01embed model - not applicable查看
WINOGRANDE0%2023-11-01embed model - not applicable查看

定价

项目价格币种
输入$0.1 / MtokUSD
输出$0.1 / MtokUSD
缓存读$0 / MtokUSD
缓存写$0 / MtokUSD

数据来源: https://cohere.com/pricing · 截至 2023-11-01

合规性

  • 数据驻留: US
  • SOC2: ✓
  • HIPAA: ✗
  • GDPR: ✓
  • ISO 27001: ✓

Embed English v3

模型概述

Cohere Embed English v3 英文嵌入模型, 512 token 输入, 为检索/分类/聚类优化, 英文语义搜索领先。

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
Cohereembed-english-v32023-11-01512textCC-BY-NC-4.0

基准测试表现

基准得分单位备注
MMLU (Massive Multitask Language Understanding)0.0%embed model - not applicable
HumanEval0.0pass@1embed model - not applicable
GSM8K (Grade School Math 8K)0.0%embed model - not applicable
MATH0.0%embed model - not applicable
BBH (BIG-Bench Hard)0.0%embed model - not applicable
GPQA0.0%embed model - not applicable
IFEval0.0%embed model - not applicable
ARC0.0%embed model - not applicable
MUSR0.0%embed model - not applicable
WinoGrande0.0%embed model - not applicable

定价

输入输出缓存读取缓存写入

每百万 token

优势

  • 可靠的通用模型。

劣势

  • MMLU 仅 0.0,知识推理偏弱。
  • HumanEval 0.0,代码能力较弱。
  • 闭源专有模型,不支持自托管。
  • 上下文窗口 0K 偏小。

适用场景

  • 通用对话与问答

参考文献