OpenAI’s latest machine-learning mode
has arrived. On Friday, OpenAI announced that its latest machine learning mode has arrived. Now you can try the o3 mini model. OpenAI makes one of its “reasoning” model available to ChatGPT users for free. To get started, click the “Reason” icon under the message creator.
OpenAI claims that o3 mini is faster and more precise than its predecessor o1 mini. The company found that o3-mini delivered a response 24 percent faster than its predecessor, o1. The new model, when set to its “medium” reason effort, can even come close to the performance the more expensive o1 in some math, programming and science benchmarks. Like OpenAI’s reasoning models, the o3 mini will explain how it came to its answer rather than simply responding to a question. The model is compatible with ChatGPT search out of the box, which allows it to search the web for useful links and the latest information. OpenAI is working to integrate search into all of its reasoning model. OpenAI said. “This model continues our track record of driving down the cost of intelligence — reducing per-token pricing by 95% since launching GPT-4 — while maintaining top-tier reasoning capabilities. As AI adoption expands, we remain committed to leading at the frontier, building models that balance intelligence, efficiency, and safety at scale.”
As of today’s announcement, the model picker will now display o3 mini instead of o1 mini. OpenAI will also triple the rate limit of Plus and Team ChatGPT from 50 messages a day with o1 mini to 150 messages a day for o3 Mini. OpenAI’s $200 per month Pro tier gives unlimited access to this new system. OpenAI’s CEO Sam Altman first announced o3 at the end last year. He said that the o3 mini would be available “around the end January.” On January 17, Altman wrote on X that OpenAI “plans to ship in a few weeks.” DeepSeek quietly launched its R1 chain of thought model on January 20, the day Altman attended Donald Trump’s inaugural. After going viral, the company’s ChatGPT app was the most downloaded free app in the US App Store by January 27. OpenAI was almost certainly blindsided by the overnight success of DeepSeek, which wiped out $1 trillion in stock market value.
After last week, OpenAI announced that it was working with Microsoft in order to identify two accounts which the company claimed may have distilled their models. Distillation is a process that transfers the knowledge from an advanced AI system into a smaller and more efficient AI system. Distillation is a common practice. DeepSeek used distillation to train its R1 model. In fact, OpenAI’s Terms of Service allow distillation so long as users do not train competing models using the AI outputs from the company. OpenAI did mention DeepSeek, but not explicitly. OpenAI’s spokesperson told The Guardian that “we know [China]-based companies — as well as others — are constantly trying distill the models of US AI companies.” David Sacks was President Trump’s AI adviser, and he was more direct. He claimed that there was “substantial proof” that DeepSeek “distilled the knowledge from OpenAI’s model.”
