Key Specifications

Vendordeepseek
Versionllm-67b
Release Date2024-01-05
Context Window32768 tokens
Input Modalitiestext
Output Modalitiestext
LicenseDeepSeek License
Documentationhttps://api-docs.deepseek.com/

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU75.2%2024-01-055-shotview
HUMANEVAL68.3pass@12024-01-05view
GSM8K79.9%2024-01-050-shot CoTview
MATH39.3%2024-01-050-shot CoTview
BBH72.6%2024-01-053-shot CoTview
GPQA37.3%2024-01-050-shotview
IFEVAL76.8%2024-01-05prompt_strictview
ARC94%2024-01-05challengeview
MUSR51.9%2024-01-050-shotview
WINOGRANDE78.1%2024-01-050-shotview

Pricing

TierPriceCurrency
Input$0.4 / MtokUSD
Output$0.4 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://api-docs.deepseek.com/quick_start/pricing · as of 2024-01-05

Compliance

  • Data Residency: CN
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

DeepSeek LLM 67B

模型概述

DeepSeek LLM 67B 通用开源模型, 32K 上下文, 中英文能力平衡, 在开源 67B 模型中表现突出。

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
Deepseekllm-67b2024-01-0532KtexttextDeepSeek License

基准测试表现

基准得分单位备注
MMLU (Massive Multitask Language Understanding)75.2%5-shot
HumanEval68.3pass@1
GSM8K (Grade School Math 8K)79.9%0-shot CoT
MATH39.3%0-shot CoT
BBH (BIG-Bench Hard)72.6%3-shot CoT
GPQA37.3%0-shot
IFEval76.8%prompt_strict
ARC94.0%challenge
MUSR51.9%0-shot
WinoGrande78.1%0-shot

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

  • 闭源专有模型,不支持自托管。

适用场景

  • 通用对话与问答

参考文献