OpenAI has announced the release of two new artificial intelligence reasoning models, known as o3 and o4-mini. The o3 model is touted as the company’s most powerful reasoning model, while the o4-mini is a smaller and faster version that offers remarkable performance for its size and cost. Both models have the capability to integrate images into their reasoning process, allowing them to “think” with visual data such as sketches and whiteboards. Additionally, these models will be equipped with all ChatGPT tools, including web browsing and image generation. Users of ChatGPT Plus, Pro, and Team can access these features immediately, with plans for broader access to come in the following weeks.
OpenAI also launched its new AI model, GPT-4.1, but notably did not release a safety report, which is typically provided with new model releases. This omission has raised concerns in the AI community, as these reports are crucial for transparency and safety evaluations. According to Steven Adler, a former OpenAI safety researcher, safety reports are voluntary but have been positioned as essential for accountability. The lack of a system card for GPT-4.1 comes amid criticisms of OpenAI’s safety practices, particularly as employees express worries about resource allocation for safety testing. Despite not being the highest-performing model, GPT-4.1 reportedly shows improvements in efficiency and response time.
OpenAI is set to phase out its GPT-4 model from ChatGPT by April 30, 2025, replacing it with the newer GPT-4o, which has shown superior performance in various evaluations, including writing and problem-solving. The company asserts that GPT-4o outperforms its predecessor consistently, thanks to recent upgrades that enhance instruction following and conversational flow.
Why do we care?
OpenAI’s release of the o3 and o4-mini reasoning models and the transition to GPT-4o in ChatGPT marks a meaningful evolution, not just in model capability, but in how AI systems are operationalized in everyday tools.
But the headline that should matter just as much—if not more—for IT services firms? There’s no safety report for GPT-4.1, and that signals deeper transparency issues even as AI becomes more embedded in client-facing tools.
On yesterday’s live show (dropping in the podcast feed this weekend, available now on YouTube), we discussed how accountability is the key to AI success in an organization. To have that accountability, one needs transparency.
Smart providers will double down on governance. Track models. Document behaviors. Insist on clarity from vendors. Because when “thinking” machines are making decisions for your clients, you’d better know how—and why—they’re reasoning the way they do.