August 7, 2025 — OpenAI officially introduced its latest flagship artificial intelligence model, GPT-5, available starting today to ChatGPT users and developers. This model has been fully upgraded in reasoning, coding, and creative generation, with OpenAI calling it one of the strongest general AI models to date.
Updates
ChatGPT Plus users can choose between GPT-5 Auto, Fast, or Thinking models. The 4o model has been re-opened to paid users, and the quota for the Thinking model is 3,000 times per week.
2025.08.12 by Sam Altman @ x.com
In ChatGPT Plus, the context window for the GPT-5 Thinking model is 196k.
2025.08.12 by Yann Dubois @ x.com
Highlights Overview
Multiple Versions Available
- GPT-5 (Standard version)
- GPT-5 mini (Lightweight version, lower cost)
- GPT-5 nano (Ultra-fast, low latency)
An automatic routing system will intelligently switch models based on task complexity.
Drastically Improved Reasoning and Coding
- SWE-Bench test score of 74.9%, and Aider Polyglot reaching 88%.
- Generate a complete webpage, game, or application prototype with a single instruction.
Stronger Multimodal Capabilities
Supports a larger context window, text and visual inputs, and higher output efficiency. In the API, all GPT-5 models can accept up to 272,000 input tokens and generate up to 128,000 reasoning and output tokens, with a total context length of 400,000 tokens.Full Ecosystem Integration
In addition to OpenAI's own platforms, it has been integrated into platforms like Microsoft 365 Copilot, GitHub Copilot, VS Code, and Azure AI Foundry.
Covering approximately 700 million ChatGPT users worldwide.
API Pricing
Under the Standard mode, the price per 1 million tokens is:
| Model | Input | Cached input | Output |
|---|---|---|---|
| gpt-5 | $1.25 | $0.125 | $10.00 |
| gpt-5-mini | $0.25 | $0.025 | $2.00 |
| gpt-5-nano | $0.05 | $0.005 | $0.40 |
- In Priority mode (faster response, higher priority), the price doubles.
- In Flex mode (higher latency), the price is halved.
Official Stance
- OpenAI CEO Sam Altman described GPT-5 as "like talking to a PhD-level expert," noting that after experiencing it, "it is hard to go back to old models."
- Despite the significant performance boost, OpenAI stated that GPT-5 is not yet AGI and lacks autonomous learning capabilities.
- Continuous improvements in safety and alignment are being made to reduce hallucinatory outputs.
Release Significance and Future Outlook
For developers, its automatic model routing and multiple API specifications bring lower costs and higher intelligence, meaning the barrier to innovation has been significantly lowered; for enterprises, it will deeply empower product R&D, customer service, and strategic decision-making; for general users, it will bring a personalized intelligence experience far beyond imagination.
Perhaps the advent of GPT-5 will be an important milestone in the history of AI development, signaling that AI is transitioning from an "information-providing tool" to a "problem-solving partner."
Source: openai.com - Introducing GPT-5
Comments