Not to be left behind in the Generative AI race, OpenAI, which caught the imagination of the world by its pathbreaking ChatGPT in 2022, has launched its largest and best model GPT 4.5 yet.

GPT‑4.5 improves its ability to recognise patterns, draw connections, and generate creative insights without reasoning by scaling unsupervised learning.

“Early testing shows that interacting with GPT‑4.5 feels more natural. Its broader knowledge base, improved ability to follow user intent, and greater Emotional Quotient make it useful for tasks like improving writing, programming, and solving practical problems,” the US-based firm said.

The best part about the new edition is that it hallucinates less. As against 80 per cent in its o3 mini, hallucination in the new version is only 37 per cent. Hallucination is an attribute of an LLM (Large Language Model such as ChatGPT and Gemini) that tells about the machine’s inability to distinguish between two or more things or persons with the same name.)

The firm said it is still exploring what it’s capable of. “We are eager to see how people use it in ways we might not have expected,” it said.

• Also read: Cognizant close to completing sale of Thoraipakkam campus

The week also saw the release of Grok 3 by xAI, the artificial intelligence firm owned by Elon Musk. Grok 3 was touted to be the smartest of all the LLMs around.

The last four weeks witnessed the rapid transformation of the LLM space with the launching of DeepSeekR1, which promised to change the way LLMs are built. With just $6 million spent on training costs, the open-source model showed the way for the countries in the Global South on how to join the LLM party, which is so far considered to be a rich countries’ game.

GPT 4.5 features

The company said unsupervised learning ability of the new model increases model accuracy and intuition. “GPT‑4.5 is an example of scaling unsupervised learning by scaling up compute and data, along with architecture and optimisation innovations,” it said.

GPT‑4.5 was trained on Microsoft Azure AI supercomputers. The result is a model that has broader knowledge and a deeper understanding of the world, leading to reduced hallucinations and more reliability across a wide range of topics.

“Combining deep understanding of the world with improved collaboration results in a model that integrates ideas naturally in warm and intuitive conversations that are more attuned to human collaboration,” it pointed out.

Access

Beginning today, ChatGPT Pro users will be able to access GPT‑4.5 in the model picker on web, mobile, and desktop. Plus and Team users will get access to it next week, then to Enterprise and Edu users the following week.