The **GPT-5 family** provides state-of-the-art performance, outperforming previous models in reasoning, coding, and factual accuracy. It features smart routing for fast responses or deep reasoning.
- **gpt-5**: Main model with advanced reasoning.
- **gpt-5-mini**: Smaller, faster, cost-effective.
- **gpt-5-nano**: Most economical option.
- **GPT-5 Pro**: Extended reasoning (ChatGPT Pro only).
- **74.9% on SWE-bench Verified**.
- **45% fewer factual errors** than GPT-4o.
- **80% fewer hallucinations** in reasoning mode.
- Pricing starts at **$0.05/1M input tokens** (gpt-5-nano).
馃敆 Full GPT-5 Documentation
GPT-5-Codex
Specialized for agentic software engineering. Can work independently for **7+ hours** on complex tasks. Available via **Codex CLI** and IDE extensions.
Model Docs
Image Generation (GPT-image-1)
Latest image model with major improvements over DALL-E 3, including better adherence to instructions, reliable text rendering, and **image-as-input** support.
Image Guide
GPT-4 Family
Proven reliability with GPT-4o (speed, multimodal) and GPT-4.1 (1M context limit). Ideal for applications requiring strong **multimodal understanding**.
Audio Models
**Whisper** (Speech-to-text), **TTS** (Text-to-Speech), and **gpt-realtime-mini** for low-latency streaming voice interactions.
Audio API Ref
