原文 · 保持原样，未翻译原文 · 未翻译
Overview
GPT-5 is OpenAI’s most intelligent model yet, and it brings major gains across reasoning, reliability, and specialized use cases.
- Dynamic reasoning (“think hard about this”) enables expert-level responses on complex tasks
- Much better at complex front-end generation
- Best model yet for health-related questions
- Significant reductions in hallucinations (~80% less with thinking mode vs. o3)
It supports a 400,000 token context window and up to 128,000 output tokens
Earlier iterations of this model were available as Horizon Alpha and Horizon Beta
SWE-Bench Verified (500 turns)
| Model | Accuracy |
|---|---|
| Open AI GPT-5 | 74.9% |
| Claude Sonnet-4 | 70.4% |
| Qwen3-Coder | 69.6% |
| OpenAI o3 | 69.1% |
Model Variants
| Model | Best for |
|---|---|
| GPT-5 | Complex reasoning, broad world knowledge, and code-heavy tasks |
| GPT-5-mini | Balanced performance: cost-optimized chat, decent reasoning, fast speed |
| GPT-5-nano | High-throughput use cases like classification or simple Q&A |
Pricing
Prices shown per 1M tokens:
- Input: $1.25
- Output: $10.00
- Input: $0.25
- Output: $2.00
- Input: $0.05
- Output: $0.40
- Input: $1.25
- Output: $10.00
Quick Links
- Check out our X post
- Official blog post from OpenAI: Introducing GPT-5
Note:
- OpenAI requires bringing your own API key to use GPT-5 over the API. Set up here: Integrations
- The free Horizon Alpha and Beta endpoints are now retired.