ChatGPT Losing Ground: The Counterattack Has Begun
Do you feel like ChatGPT isn't performing as well as it used to? You're not alone. Users are migrating to more powerful alternatives, and OpenAI knows it.
It's in this context of market share loss that GPT-5.4 arrives. A major update, according to OpenAI, but is it enough to reverse the trend?
GPT-5.4 Thinking: The Model That Wants to Do Everything
GPT-5.4 comes in several flavors. The flagship version, "Thinking," promises advanced reasoning and optimized tool management. Imagine a virtual assistant capable of not only understanding your requests, but also interacting directly with your computer.
The native integration of "computer use" capabilities is a real breakthrough. No more complex plugins, GPT-5.4 can manipulate your interface like a human user.
A Work Plan for More Control
Another interesting feature: the display of a work plan upstream for complex requests. This feature allows you to adjust the instructions along the way, avoiding endless and frustrating iterations. A considerable gain in time and efficiency.
ChatGPT for Excel: AI at the Heart of Your Spreadsheets
The integration of ChatGPT into Excel is another significant advance. ChatGPT for Excel allows you to build, analyze, and update models in natural language, directly within your spreadsheets. Forget complex formulas, AI takes care of everything.
Currently reserved for Business, Enterprise, Edu, Pro, and Plus subscribers in the United States, Canada, and Australia, this feature promises to transform the way businesses use Excel.
Hallucinations: Does the Weak Point Persist?
OpenAI promises a reduction in hallucinations with GPT-5.4. Fewer factual errors, more reliable answers... But should we take their word for it?
The crisis of credibility of standardized AI model evaluations is a real problem. OpenAI's internal benchmarks should be taken with caution. Only real-world use will verify the effectiveness of this correction.
The AI Race is On: Who Will Win?
GPT-5.4 is a clear response to competitive pressure. Anthropic and other players are gaining ground, and OpenAI must react.
The question isn't whether GPT-5.4 is a good model. The question is whether it's *good enough* to stop the bleeding and give OpenAI its initial lead back. At Dumont Consulting, we believe the battle has only just begun. Businesses must remain attentive to developments and choose the tools best suited to their needs, without being blinded by marketing promises.
The AI market is constantly evolving. GPT-5.4 is a step, but not the final destination. The future will tell us who wins this frantic race.