Benchmark Performance
| Benchmark |
Score |
Unit |
Evaluated At |
Notes |
Source |
| MMLU |
50.3 |
% |
2022-11-30 |
5-shot |
view |
| HUMANEVAL |
28.2 |
pass@1 |
2022-11-30 |
— |
view |
| GSM8K |
56.6 |
% |
2022-11-30 |
0-shot CoT |
view |
| MATH |
29.4 |
% |
2022-11-30 |
0-shot CoT |
view |
| BBH |
63.9 |
% |
2022-11-30 |
3-shot CoT |
view |
| GPQA |
26.9 |
% |
2022-11-30 |
0-shot |
view |
| IFEVAL |
52.8 |
% |
2022-11-30 |
prompt_strict |
view |
| ARC |
81.8 |
% |
2022-11-30 |
challenge |
view |
| MUSR |
40.4 |
% |
2022-11-30 |
0-shot |
view |
| WINOGRANDE |
72.7 |
% |
2022-11-30 |
0-shot |
view |
Pricing
| Tier |
Price |
Currency |
| Input | $2 / Mtok | USD |
| Output | $2 / Mtok | USD |
| Cache Read | $0 / Mtok | USD |
| Cache Write | $0 / Mtok | USD |
Source:
https://openai.com/api/pricing/
· as of 2022-11-30
Compliance
- Data Residency: US
- SOC2: ✓
- HIPAA: ✓
- GDPR: ✓
- ISO 27001: ✓
GPT-3.5
Model Overview
OpenAI GPT-3.5 基础模型, 4K 上下文, ChatGPT 初始版本所用模型, 推理能力优于 GPT-3。
Core Specifications
| Vendor |
Version |
Release Date |
Context Window |
Input Modalities |
Output Modalities |
License |
| Openai |
3.5 |
2022-11-30 |
4K |
text |
text |
Proprietary |
| Benchmark |
Score |
Unit |
Notes |
| MMLU (Massive Multitask Language Understanding) |
50.3 |
% |
5-shot |
| HumanEval |
28.2 |
pass@1 |
— |
| GSM8K (Grade School Math 8K) |
56.6 |
% |
0-shot CoT |
| MATH |
29.4 |
% |
0-shot CoT |
| BBH (BIG-Bench Hard) |
63.9 |
% |
3-shot CoT |
| GPQA |
26.9 |
% |
0-shot |
| IFEval |
52.8 |
% |
prompt_strict |
| ARC |
81.8 |
% |
challenge |
| MUSR |
40.4 |
% |
0-shot |
| WinoGrande |
72.7 |
% |
0-shot |
Pricing
| Input |
Output |
Cache Read |
Cache Write |
| — |
— |
— |
— |
per million tokens
Strengths
- Reliable general-purpose model.
Weaknesses
- MMLU 50.3, weak knowledge reasoning.
- HumanEval 28.2, coding weak.
- Proprietary, not self-hostable.
- Context window 4K is limited.
Use Cases
References