OpenAI has officially launched GPT-5, the latest version of its prominent language model, making it available for both free and paying ChatGPT users. During a recent press briefing, CEO Sam Altman described GPT-5 as a significant advancement towards achieving artificial general intelligence (AGI), although he noted that it has not yet reached this milestone. Specifically, he emphasized that while GPT-5 exhibits intelligence comparable to that of a human expert, it still lacks certain key capabilities associated with AGI, such as continuous learning post-deployment.
OpenAI claims that GPT-5 is more intelligent, faster, and accurate than its predecessors, with a significantly reduced hallucination rate. Altman compared the advancements from GPT-4 to GPT-5 to the evolution from pixelated displays to Retina screens, stating, “GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.”
The launch also introduced two new model variants: a lightweight GPT-5-mini and a faster, more economical GPT-5-nano, with the latter currently only available through the API. Free users will have access to both GPT-5 and GPT-5-mini, while the Plus subscription will include increased usage limits. The $200 per month Pro tier grants unlimited access to GPT-5 and a more powerful version called GPT-5-pro, as well as GPT-5-thinking, which extends processing times for complex queries. The chat interface will intelligently route queries to appropriate models based on user needs and subscription level.
Developers using the API can expect pricing at $1.25 per million input tokens and $10 per million output tokens for GPT-5. For GPT-5-mini, the cost is $0.25/1M input and $2/1M output tokens, while GPT-5-nano is priced at $0.05/1M input and $0.40/1M output tokens, making it an attractive option for developers who previously utilized more cost-effective models.
Starting soon, Pro users will have the ability to sync their Gmail, Google Contacts, and Google Calendar with ChatGPT, with plans to extend this feature to other subscription tiers later. OpenAI is also rolling out customization features, allowing users to select a chat color and choose from four preset personalities: Cynic, Robot, Listener, or Nerd.
Moreover, GPT-5 boasts a 256,000-token context window—up from the previous model’s 200,000 tokens—enhancing its capability to manage lengthy conversations, documents, or code without losing context. Performance benchmarks indicate that GPT-5 outperforms earlier models in several coding assessments.
During demonstrations, GPT-5 showed remarkable performance, crafting a fully interactive web application in a matter of minutes based on user-provided instructions. Additionally, OpenAI indicates that GPT-5 demonstrates improved reliability for health-related inquiries, achieving substantial gains in scores across health benchmarks compared to older versions.
OpenAI reports that they have significantly reduced the incidence of hallucinations—a common issue where AI generates inaccurate information. The GPT-5 model is said to have a 26% lower hallucination rate than the GPT-4o model. Following rigorous testing, it has been confirmed that GPT-5-thinking has been trained to respond safely to prompts requiring careful handling.
The company also noted that it serves around 700 million weekly active ChatGPT users, alongside 5 million paying business customers and 4 million developers leveraging the API. Nick Turley, head of ChatGPT, expressed confidence that users, particularly those who are less technically minded, will find the enhanced capabilities of GPT-5 notably beneficial.
