Key Specifications
| Vendor | other |
| Version | dbrx-instruct |
| Release Date | 2024-03-27 |
| Context Window | 32768 tokens |
| Input Modalities | text |
| Output Modalities | text |
| License | DBRX Open Model License |
| Documentation | https://huggingface.co/models |
Benchmark Performance
| Benchmark |
Score |
Unit |
Evaluated At |
Notes |
Source |
| MMLU |
81.3 |
% |
2024-03-27 |
5-shot |
view |
| HUMANEVAL |
72.8 |
pass@1 |
2024-03-27 |
— |
view |
| GSM8K |
78.7 |
% |
2024-03-27 |
0-shot CoT |
view |
| MATH |
50.1 |
% |
2024-03-27 |
0-shot CoT |
view |
| BBH |
71.7 |
% |
2024-03-27 |
3-shot CoT |
view |
| GPQA |
34.7 |
% |
2024-03-27 |
0-shot |
view |
| IFEVAL |
73.1 |
% |
2024-03-27 |
prompt_strict |
view |
| ARC |
93.9 |
% |
2024-03-27 |
challenge |
view |
| MUSR |
54.7 |
% |
2024-03-27 |
0-shot |
view |
| WINOGRANDE |
81.8 |
% |
2024-03-27 |
0-shot |
view |
Pricing
| Tier |
Price |
Currency |
| Input | $0.75 / Mtok | USD |
| Output | $0.75 / Mtok | USD |
| Cache Read | $0 / Mtok | USD |
| Cache Write | $0 / Mtok | USD |
Source:
https://huggingface.co/models
· as of 2024-03-27
Compliance
- Data Residency: self-host
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
DBRX Instruct
Model Overview
Databricks DBRX Instruct MoE 模型, 32K 上下文, 总参 132B/活跃 36B, 在开源模型中编程/数学领先。
Core Specifications
| Vendor |
Version |
Release Date |
Context Window |
Input Modalities |
Output Modalities |
License |
| Other |
dbrx-instruct |
2024-03-27 |
32K |
text |
text |
DBRX Open Model License |
| Benchmark |
Score |
Unit |
Notes |
| MMLU (Massive Multitask Language Understanding) |
81.3 |
% |
5-shot |
| HumanEval |
72.8 |
pass@1 |
— |
| GSM8K (Grade School Math 8K) |
78.7 |
% |
0-shot CoT |
| MATH |
50.1 |
% |
0-shot CoT |
| BBH (BIG-Bench Hard) |
71.7 |
% |
3-shot CoT |
| GPQA |
34.7 |
% |
0-shot |
| IFEval |
73.1 |
% |
prompt_strict |
| ARC |
93.9 |
% |
challenge |
| MUSR |
54.7 |
% |
0-shot |
| WinoGrande |
81.8 |
% |
0-shot |
Pricing
| Input |
Output |
Cache Read |
Cache Write |
| — |
— |
— |
— |
per million tokens
Strengths
- MMLU score 81.3, strong knowledge reasoning.
- Mixture-of-Experts architecture.
Weaknesses
- Proprietary, not self-hostable.
Use Cases
- Code generation and debugging
References