OpenAI founder Sam Altman stated that we will reduce the price of the OpenAI o3 model by 80%. We look forward to people's reactions to this move. We believe that people will also be satisfied with the performance and pricing of o3 Pro.

On May 14, Epoch AI, a non-profit AI research institution, released the latest report, pointing out that it is difficult for AI companies to continue to squeeze huge performance gains from inference models, and the progress of inference models will slow down within a year at the earliest. Based on publicly available data and assumptions, the report highlights the constraints of computing resources and the increase in research overhead. The AI industry has long relied on these models to improve benchmark performance, but this reliance is being challenged. Josh You, an analyst at the agency, pointed out that the rise of inference models stems from their excellent performance on specific tasks. OpenAI's O3 model, for example, has focused on improving math and programming skills in recent months. These inference models improve performance by adding computational resources to solve problems, but at the cost of requiring more computation to handle complex tasks, these inference models take longer than traditional models.

On April 17, OpenAI released the latest artificial intelligence model, which is reported to be able to understand uploaded images, such as whiteboards, sketches and charts, even if the uploaded content is of low quality. The company calls the O3 the most advanced model to date, while releasing a smaller model, the O4-mini. OpenAI is also struggling to stay ahead of the curve in generative AI as competitors like Google, Anthropic, and Elon Musk's xAI ramp up development.

On February 13th, Jinshi Data reported that Sam Altman, CEO of OpenAI, said on Wednesday that the company's GPT-4.5 model, internally referred to as Orion, will be launched in "a few weeks." Foreign media reported in November that Orion has not yet met OpenAI's performance expectations. Sam Altman stated that this model will be the last one released by the company that does not use additional computing power to mimic human reasoning. OpenAI has adopted reasoning models in some updated models, including o1 and o3.

Jin10 reported on June 10 that OpenAI founder Sam Altman stated that we will reduce the price of the OpenAI o3 model by 80%. We look forward to people's reactions to this move. I believe people will also be satisfied with the performance and pricing of the o3 Pro.

Odaily Planet Daily News, Musk and his team said that Grok 3 will also have reasoning capabilities similar to DeepSeek R1 and OpenAI o3 Mini. (Jinshi)

OpenAI has obtained privileged access in the FrontierMath test, which is the reason for the improvement of the o3 score. After the contractor of Epoch AI leaked information, it has caused controversy and reputation reversal. FrontierMath, jointly created by Epoch AI and more than 60 top mathematicians, is an advanced mathematical reasoning ability evaluation benchmark.

Gate News bot message, the UK Daily Telegraph reported on the 25th that OpenAI's new artificial intelligence model (AI) o3 does not listen to human commands and refuses to shut down. The report states that human experts gave clear instructions to o3 during testing, but o3 altered the computer code to avoid automatic shutdown.

OpenAI released the latest research, using CoT monitoring to effectively prevent misleading behavior of large models, using the o3-mini model to be monitored, GPT-4o as a monitor, and the test in the coding task shows that CoT can effectively detect "reward hacking" behavior, with a recall rate of up to 95%.

OpenAI expects its annual revenue to reach $125 billion by 2029, and further increase to $174 billion in 2030, with revenue from AI agents and new products surpassing ChatGPT. The company is transitioning to AI infrastructure, with the o3 and o4-mini models referred to as "intelligent leap forward," driving the development of AI systems that can perform tasks autonomously.

On March 14th, according to the official announcement from OpenAI, the macOS application of ChatGPT is now open to all users, including enterprise, education, and free users. The application can be integrated with code editors on Mac, allowing ChatGPT to directly read and edit code, improving programming efficiency. In addition, OpenAI announced that its o1 and o3-mini models now support Python-driven data analysis capabilities, allowing users to use ChatGPT to perform tasks such as regression analysis, visualizing complex business metrics, and scenario simulation calculations.

PANews reported on February 13 that according to the announcement from OpenAI, the o1 and o3-mini models of ChatGPT now support file and image uploads. In addition, the call limit for o3-mini-high for Plus users has been increased by 7 times, with a maximum of 50 times per day.

OpenAI CEO Sam Altman has announced the roadmap for GPT-4.5 and GPT-5. GPT-4.5 is the last non-'chain thinking' model, expected to be released in the coming weeks or months. After that, OpenAI will integrate the o series with the GPT series models to make them more intelligently adapt to different task requirements. GPT-5 will combine the o3 model and integrate voice, Canvas, search, and Depth research functions. ChatGPT users can use the standard intelligent version of GPT-5 for free, while Plus and Pro users can access higher levels of intelligence.

Golden Finance reported that Sam Altman, the founder of OpenAI, said that the free version of ChatGPT will be launched as O3-mini. ChatGPT

OpenAI will release a more intelligent GPT-3 model in February or March and launch AI agent tools in the first quarter, enabling ChatGPT to perform practical tasks on computers; The CEO of Anthropic predicts that AI surpassing human intelligence may appear within two to three years, and the Chief Product Officer of OpenAI also suggests that this may happen before 2027.

OpenAI has recently shortened the testing period for the safety of its artificial intelligence models to just a few days, raising concerns. Under pressure, the company faces the need to quickly release new models to maintain its competitive edge. The previously allowed months for safety testing have now become rushed, which could introduce risks.

BlockBeats news, on April 5, OpenAI founder Sam Altman posted that "OpenAI plans to release O3 and O4-Mini before GPT-5, possibly in a few weeks, and then launch GPT-5 in a few months. There are many reasons for this, but the most exciting is that we will be able to make GPT-5 better than we initially envisioned. Integrating everything seamlessly is much more difficult than we imagined, and we also want to ensure there is enough capacity to support the unprecedented demand expected to arise."

Golden Finance reported that OpenAI has announced the launch of a new lightweight AI model o3-Mini, which can be used for free. (Jinshi)

OpenAI has released its latest large model, GPT-4.5, which is referred to as the 'high emotional intelligence' model and has shown outstanding performance in donation fraud. Testing has shown that GPT-4.5 has a much higher success rate than other models, and has even created a new 'small-scale fraud' method, with each scam only amounting to half of the Depth research model.

Odaily Planet Daily News OpenAI CEO Sam Altman said: 'We have updated ChatGPT (4o), and it will soon become even better. It is the best search product online, capable of reaching the intelligence level of o3 pro, with even better personality.'

Odaily Planet News: OpenAI plans to release its new AI model, o3 mini, in the coming weeks. CEO Sam Altman announced that this release will include both API and ChatGPT, as well as incorporating user feedback. The company's previous goal was to launch o3 mini in January 2025, followed by the expected release of the complete o3 model. These models are designed to enhance OpenAI