OpenAI rolls out o3-mini to rival DeepSeek's R1
More than just an incremental upgrade, o3-mini is a clear signal to DeepSeek that OpenAI isn’t ceding ground easily.
For the first time in a long while, OpenAI found itself in an unusual position—on the defensive. Last week, Chinese AI company DeepSeek surged onto the scene with its R1 model that rattled the entire industry.
OpenAI had alleged that DeepSeek’s sudden rise wasn’t entirely organic and may have involved stolen intellectual property. But, whether the claims are true or not, DeepSeek is shaping up to be the most serious challenger to OpenAI’s dominance yet.
Not willing to take the hit lying down, OpenAI struck back by Friday. Its answer? o3-mini—a smarter, faster, and more cost-efficient model, directly challenging DeepSeek R1's biggest selling point – all completely free. No subscription, no paywall—just OpenAI putting its best foot forward at a critical moment.
Oyinebiladou Omemu
First teased in December, OpenAI brags that o3-mini "advances the boundaries of what's possible with smaller, more efficient models. Like September's o1-mini before it, the model is optimized for STEM applications, excelling in science, math, and coding" despite lower operating costs and latency than o1-mini.
And OpenAI has the numbers to back it up. According to the company's internal tests, o3-mini is 39% less prone to major errors than o1-mini while delivering responses 24% faster. It’s also the first small reasoning model from OpenAI to support advanced features like function calling, structured outputs, and developer messages, making it more adaptable for complex tasks.
Emmanuel Oyedeji
It features additional flexibility, allowing users to choose from three reasoning effort levels—low, medium, and high—giving them control over the balance between speed and accuracy. OpenAI also integrated a search function, allowing the model to find up-to-date answers with linked sources for real-time information retrieval.
But, perhaps, the most surprising aspect of the release is who gets access. For the first time, OpenAI is making a reasoning model available to free users. Anyone using ChatGPT can now select the “Reason” option and try it out. Paid subscribers on the Plus, Team, and Pro tiers will see o3-mini replace o1-mini in their model picker, with a significant bump in usage limits—up to 150 messages per day, up from the previous 50.
If the timing of this release says anything, it’s that OpenAI is sending a deliberate and clear signal to DeepSeek that it isn’t ceding ground anytime soon. Interestingly, Alibaba wants to send a similar message with its Qwen 2.5 Max launched last week. It would be interesting how the reignited AI rivalry plays out over the coming weeks.
Kelechi Edeh
