Here’s everything new and different in the GPT-5 (Openai’s long-awaited flagship AI model)



Openai releases new LLM GPT-5. The new Generative AI model will be available to consumers for free and power the latest version of the popular Chatgpt Bot, which represents “our smartest, fastest, and most useful model with built-in thinking that takes expert-level intelligence into everyone’s hands,” Openai said Thursday.

If you’re wondering what all Hoopla means, and what’s better or different from the company’s previous GPT models (or competitor AI models like Claude, Gemini, or Llama), here’s a quick record of some of the most important new features available in GPT-5:

Easy to use

The latest version of OpenAI products forces users to choose the type of model they want to use for different tasks – OpenAI’s O series of “inference” models, for complex research, or standard GPT for rapid results. GPT-5 uses a “real-time router” that automates the process and selects the right tool for the right job, so you don’t have to.

Special characters

Chatgpt will now let users choose from four different preset “personalities”: “cynical”, “robot”, “listener” and “nerd”. These personalities are designed to make using chatbots more natural and context-friendly. If you want more irony, choose cynical people, and if you use it for work, a “efficient and blunt” robot character might be better.

Less hallucinations

Openai claims that the GPT-5 is not easy to invent information or hallucinate. According to OpenAI, GPT-5 is 45% less likely to contain factual errors than GPT-4O in tests that enable web search and use anonymous tips. When “thinking”, GPT-5 may have 80% fewer errors than OpenAI O3. But this still doesn’t eliminate the hallucinations, nor does it Mashable Explainwhich means that the GPT-5 will still hallucinate every ten times on normal tasks.

Better writers

According to Openai, the GPT-5 is better than its predecessors, whose “engaging, resonant writing with literary depth and rhythm.” The company offers a side-by-side comparison of GPT-5’s brushwork with things like wedding toast and poetry.

Better encoder

While previous versions of GPT had software encoding capabilities, OpenAI said GPT-5 brought improved and sophisticated features to its “Vibe encoding” feature, allowing users to “create beautiful and responsive websites, applications, and games” in one prompt.

Agent Function

GPT-5 with Gmail and Google Calendars to assist with scheduling, reminders, email follow-ups and other productivity tasks. (Initially, this feature will be limited to users with a paid “Pro” membership by the company.)

healthy

LLM has become a popular way for people to obtain medical information and health advice. Openai said that GPT-5 scored higher in the HealthBench test than any of its previous models, and it worked more like “an active partner of thought, proactively labeling potential problems and asking questions to give more useful answers.” Of course, OpenAI also noted that it cannot replace medical professionals.

Your mileage may vary

These are just some new features and improvements that will provide consumers and businesses with plenty of opportunities to try out the latest models and compare them to the current preferred model. Whether representing a significant leap, or as external experts test models and carry them with real-world users, it will become clear in the days and weeks ahead. Early testers interviewed by Reuters explain The improvement from GPT-4 to GPT-5 is “not as big as from GPT-3 to GPT-4”. Artificial intelligence experts, usually doubt Gary Marcus says: “Fans will still find something welcoming, but the GPT-5 is not the huge leap that people have long expected.”

Introduce 2025 Fortune Global 500the world’s largest companies are certainly ranked. Explore this year’s list.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *