OpenAI launches O3-Pro to tackle complex questions: Here's how it works
OpenAI has announced a new version of its language model called O3-Pro. This model is made to better handle complex questions and follow instructions more carefully. It is now available to ChatGPT Plus users, as well as to business and API users.
O3-Pro is the first release under OpenAI's new naming system, which replaces version numbers with a monthly format. The 'O' marks it as the June 2025 release. This change supports more regular updates.
The model is part of the GPT-4 group and focuses on improving reasoning, maintaining context, and producing more consistent responses. Earlier versions sometimes struggled with multi-step logic or unclear instructions. O3-Pro addresses these issues by responding more accurately to detailed prompts.
It is now the default for ChatGPT Plus users and requires no special setup. Future updates will follow the same release approach, providing the latest version automatically.
Those who have used the model note improvements in how it handles longer conversations. It seems to better understand the intent behind complex questions, helping reduce confusion and follow up corrections.
For example, O3-Pro is better at following step-by-step instructions, such as legal queries that unfold over multiple parts, without losing track of earlier details. Previous models often required users to repeat or rephrase inputs in such scenarios.
According to OpenAI, O3-Pro was built to 'handle more complex reasoning with better consistency,' with training focused on reducing errors and improving clarity across a broader range of topics.
These updates are particularly useful in fields like research, law, and technical writing, where accuracy and deeper understanding are important.
O3-Pro can support a range of tasks, including writing, coding, reports, and data analysis. OpenAI notes that it now handles more types of input with fewer mistakes, which may help teams working with technical or structured content.
Despite improvements, OpenAI acknowledges that O3-Pro can still produce incorrect or unclear answers, especially with recent or fast changing information. Human review remains important in such cases.
The release comes at a time when other companies are also launching AI models, some with features like memory or access to live data. OpenAI continues to focus on improving general purpose tools through regular updates.
How O3-Pro performs in everyday use may influence what users expect from AI tools in the future.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
&w=3840&q=100)
Business Standard
5 hours ago
- Business Standard
Bhavish Aggarwal's Krutrim launches AI assistant Kruti for daily tasks
Krutrim, the artificial intelligence startup co-founded by Ola founder Bhavish Aggarwal, on Thursday launched Kruti, an AI assistant it says represents a leap beyond conventional chatbots. Unlike earlier tools focused on passive text replies, Kruti is positioned as an 'agentic AI'—designed to proactively carry out tasks such as cab booking, food ordering, bill payments, image generation and research. The launch builds on Krutrim's earlier public beta and introduces a redesigned interface with a focus on advanced reasoning and personalisation. 'Kruti is the first real step towards the future of AI where technology doesn't just talk back, but actually helps you get things done,' said Bhavish Aggarwal, founder of Krutrim. 'We have built Kruti to work the way Indians live: multilingual, mobile-first and intuitive.' The platform also includes features such as read-aloud responses and free access to premium AI capabilities like research and image creation. The company said Kruti aims to make advanced AI more accessible to Indian users by combining broad functionality with local context. Kruti understands both voice and text input, remembers past interactions and tailors responses in tone, length and language, including 13 Indian languages. Whether it's 'book me a cab to the airport' or 'order my usual lunch', Kruti understands intent and executes the task without micromanagement. Krutrim said Kruti is powered by its latest large language model, Krutrim V2, alongside cutting-edge open-source systems, enabling scalable and cost-efficient performance tailored for Indian users. Personalisation is at the heart of Kruti's design. It learns user preferences, adapts to usage over time and integrates with various apps and services to deliver context-aware assistance with minimal effort. Kruti also includes a fully embeddable software development kit (SDK), allowing developers to integrate large language model orchestration, memory handling and tool execution with minimal code. By offering features such as research tools, image generation and task-based agents at no cost, the company is positioning Kruti as a more accessible and affordable alternative in the AI space. Built with human-centred design, Kruti aims to eliminate app-switch fatigue and formats responses for clarity using summaries, tables and stories. It reflects how Indians use and relate to technology in everyday life. Krutrim is part of a growing ecosystem of Indian companies building AI solutions, often with a focus on local languages and data. In the Indian context, it competes with global AI giants such as OpenAI, Anthropic and Google, as well as local players such as Sarvam AI and Earlier this year, Bhavish Aggarwal announced an investment of ₹2,000 crore in Krutrim, with a commitment to invest an additional ₹10,000 crore by next year. The company also launched the Krutrim AI Lab, published several technical reports and released some of its work to the open-source community. The move followed DeepSeek's decision to open-source its own generative AI (GenAI) model. BharatGPT Mini enters the fray On Thursday, CoRover, a conversational AI firm, also launched BharatGPT Mini, a compact small language model (SLM) designed for use in low-compute and infrastructure-constrained settings. Trained on the company's proprietary conversational dataset, the 534-million parameter model supports text-based tasks in 14 Indian languages, aiming to expand AI access across underserved regions. The model is optimised for edge and server implementations, enabling seamless AI performance on low-end devices. This strategic development addresses the growing demand in India for cost-effective, fast and privacy-centric AI solutions that operate without reliance on high-end cloud infrastructure. CoRover said India's need for SLMs is accelerating, driven by demand for faster, localised and cost-efficient AI tools that maintain data privacy. While large language models (LLMs) remain essential for complex, multi-domain tasks, SLMs are increasingly preferred for their domain specificity, speed, offline functionality and ease of integration on constrained devices.
Time of India
6 hours ago
- Time of India
No new launch date for Axiom-4 mission to ISS yet
A Falcon 9 rocket stands at the pad for the launch of the Axiom-4 crewed mission to the International Space Station The Axiom-4 mission launch, including Indian astronaut Shubhanshu Shukla, has been postponed. A leak was detected in SpaceX's Falcon-9 rocket. NASA and Roscosmos are investigating a pressure signature on the International Space Station's Zvezda module. Tired of too many ads? Remove Ads Tired of too many ads? Remove Ads NASA and Axiom Space were yet to announce a new date for the launch of the Axiom-4 iss ion to the International Space Station (ISS), involving India's astronaut Shubhanshu Shukla and three others, after the lift-off was delayed due to a leak detected in SpaceX's Falcon-9 rocket."A new launch date for the fourth private astronaut mission will be provided once available," a joint statement from Axiom Space and NASA said, announcing the postponement of the launch of the Axiom-4 statement said as part of an ongoing investigation, NASA was also working with Roscosmos to understand a new pressure signature, after the recent post-repair effort in the aftmost segment of the International Space Station's Zvezda service module."Cosmonauts aboard the space station recently performed inspections of the pressurised module's interior surfaces, sealed some additional areas of interest, and measured the current leak rate. Following this effort, the segment is now holding pressure," the statement saidThe postponement of Axiom Mission 4 provides additional time for NASA and Roscosmos to evaluate the situation and determine whether any additional troubleshooting is necessary. NASA defers to Roscosmos to answer specific questions about the Zvezda module, the statement NASA astronaut and director of human spaceflight at Axiom Space Peggy Whitson will command the commercial mission, while Indian Space Research Organisation astronaut Shubhanshu Shukla will serve as two mission specialists are ESA (European Space Agency) project astronaut Sławosz Uznański-Wiśniewski of Poland and Tibor Kapu of crew will lift off aboard the SpaceX Dragon spacecraft on Falcon 9 from Launch Complex 39A at NASA's Kennedy Space Center in 14-day mission will "realise the return" to human spaceflight for India, Poland and Axiom-4 mission was originally scheduled for launch on May 29. It was first postponed to June 8 and later to June June 10, the launch was postponed by a day due to bad weather in the flight path. On June 11, SpaceX engineers detected a leak in the rocket's booster, forcing them to delay the launch till the problem was resolved.
Economic Times
6 hours ago
- Economic Times
OpenAI in talks to raise fresh funds from Saudi Arabia's PIF, India's Reliance Industries: The Information
OpenAI has held discussions with Saudi Arabia's Public Investment Fund (PIF), Reliance Industries, and UAE's MGX to participate in its ongoing $40 billion fundraising, US tech publication The Information reported. ADVERTISEMENT The talks are part of the second instalment of the fundraise, where OpenAI is looking to raise about $30 billion by December, potentially valuing the company at around $260 billion before the new investment, the report said. Japan's SoftBank, which is leading the round, has already been actively buying shares from employees. Between March and May, it purchased around $240 million worth of shares from a small group of current and former staffers, the report said. Last year, ahead of its mega round, SoftBank had also acquired about $2 billion worth of new and existing shares. SoftBank is expected to contribute at least three-fourths of the $30 billion instalment, leaving about $7.5 billion to be raised from new investors, according to the report. The Japanese conglomerate may take loans or sell some of its public market holdings, including its shares in Arm Holdings, to fund the investment, the report has also discussed raising at least $100 million each from Coatue Management and Founders Fund as part of the ongoing fundraise, the report said, while talks with PIF and other prospective investors are still at an early stage.A direct investment by Saudi Arabia's PIF, which manages over $900 billion in assets, would mark a shift from its earlier exposure to US tech firms through venture funds like Andreessen Horowitz and Iconiq Capital, The Information noted. ADVERTISEMENT CEO Sam Altman had earlier met with Saudi Crown Prince Mohammed bin Salman during a visit to Riyadh, which coincided with the kingdom's latest AI investment push, the report in India, OpenAI is exploring a product and sales partnership with Reliance, whose portfolio includes Jio, India's largest wireless carrier, and the world's biggest oil refinery. OpenAI wants Reliance businesses to distribute or sell its AI offerings in India, which is now its second-largest user market, Altman said in February, the report said. ADVERTISEMENT The scale of its capital needs is also rising sharply as ChatGPT's weekly active users exceed 500 million, up from 300 million in December. Between 2025 and 2027, OpenAI expects to spend about $35 billion on servers to support existing products, and another $55 billion on developing new models, the report of the round hinges on a pending corporate restructuring of its for-profit entity. While an earlier plan to convert it into a public benefit corporation was dropped, a revised restructuring still needs to be finalised by year-end. Failure to do so could see SoftBank cut its commitment and scale back the round to $20 billion, the report said. ADVERTISEMENT Separately, OpenAI, SoftBank, MGX and Oracle have partnered on Stargate, a $500 billion initiative to build global AI data centres, according to The Information.
