Key Specifications

Vendorother
Versiondbrx-base
Release Date2024-03-27
Context Window32768 tokens
Input Modalitiestext
Output Modalitiestext
LicenseDBRX Open Model License
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU79.9%2024-03-275-shotview
HUMANEVAL72.1pass@12024-03-27view
GSM8K75.4%2024-03-270-shot CoTview
MATH36%2024-03-270-shot CoTview
BBH72.3%2024-03-273-shot CoTview
GPQA39.3%2024-03-270-shotview
IFEVAL79.3%2024-03-27prompt_strictview
ARC93.3%2024-03-27challengeview
MUSR53.8%2024-03-270-shotview
WINOGRANDE79.8%2024-03-270-shotview

Pricing

TierPriceCurrency
Input$0.75 / MtokUSD
Output$0.75 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-03-27

Compliance

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

DBRX Base

Model Overview

Databricks DBRX Base 基础 MoE 模型, 32K 上下文, 总参 132B/活跃 36B, 适合企业定制微调。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Otherdbrx-base2024-03-2732KtexttextDBRX Open Model License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)79.9%5-shot
HumanEval72.1pass@1
GSM8K (Grade School Math 8K)75.4%0-shot CoT
MATH36.0%0-shot CoT
BBH (BIG-Bench Hard)72.3%3-shot CoT
GPQA39.3%0-shot
IFEval79.3%prompt_strict
ARC93.3%challenge
MUSR53.8%0-shot
WinoGrande79.8%0-shot

Pricing

InputOutputCache ReadCache Write

per million tokens

Strengths

  • Mixture-of-Experts architecture.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Code generation and debugging

References