In an unsurprising move for the times we live in, Microsoft has partially released its ThinkDeeper feature, which was previously exclusive to premium Copilot Pro users, to the general public.
The news, initially leaked by users on technical forums and confirmed last week, marks a milestone in the race to bring advanced artificial intelligence to the masses.
But what is behind this decision? How does ThinkDeeper actually work? And, above all, what role does OpenAI play in this technological chess game?
We are going to unravel the keys to the new functionality, as well as explore its relationship with OpenAI models, the implications for the future of AI and the limits that still persist.
When AI stops responding and starts thinking
ThinkDeeper is not a simple update: it is a paradigm shift in how we interact with AI.
While tools like Gemini or the standard Copilot prioritize quick (although sometimes superficial) answers, ThinkDeeper activates an “extended reasoning” mode that allows for more detailed and in-depth analysis.
This function decomposes complex questions into subtasks, evaluates scenarios and builds a chain of logic before providing an answer.
This makes it ideal for situations that require critical thinking, such as advanced mathematics, code debugging, strategic planning, or scientific consultation.
In addition, it adapts to the user’s implicit goals, customizing its responses according to the context.
For example, if you order an exercise routine, ThinkDeeper not only lists activities, but also analyzes your protein intake, suggests nutritional adjustments, and warns about potential mistakes.
How to activate ThinkDeeper in Copilot?
In principle all Copilot chatbot users can activate reasoning mode. To use it in a query, you just have to look for the “Think Deeper” button next to the chat box.
Once activated, a spiral icon will appear during processing. It’s worth remembering that free users are limited to three messages daily, while Pro subscribers can send up to 50 weekly, depending on demand.
The OpenAI connection: o1-mini, o3-mini or something else?
Microsoft and OpenAI have been closely collaborating for years, and ThinkDeeper is the latest fruit of this alliance. However, the exact reasoning model driving this feature remains a mystery. But speculating is free.
In September 2023, OpenAI announced O1 Strawberry, a model with “reasoning tokens” designed to solve problems step by step. A month later, Microsoft released ThinkDeeper for Copilot Pro, confirming that it used an adapted version of O1-preview.
Currently, there is speculation that the free version of ThinkDeeper could be using O1-mini, a variant optimized for specific tasks with lower resource consumption, or even O3-mini, the next OpenAI model scheduled for later this month.
Microsoft maintains secrecy about these details, probably to protect its competitive advantages and avoid excessive expectations among users.
##ThinkDeeper vs. The competition: Who is leading?
ThinkDeeper is not a general-purpose model, but rather a specialized deep analysis tool.
Unlike GPT-4o, which shines in creativity, ThinkDeeper positions itself as the “expert” for solving technical problems. Although Google has also made progress with Gemini, its focus remains limited in the area of complex reasoning.
While Copilot ThinkDeeper offers free access with restrictions and specialized reasoning, ChatGPT with o1 is only available to subscribers, and Google Gemini still doesn’t match Microsoft’s level of deep analysis.
Why is Microsoft giving away its crown jewel?
This decision is not altruistic. After analyzing usage patterns and previous strategies, three main reasons can be identified:
First, ThinkDeeper acts as a demo tool to attract users to Copilot Pro. It’s like trying out a Ferrari before you buy it: those who get used to detailed answers will hardly return to basic alternatives.
Second, releasing this feature allows Microsoft to collect valuable data on complex problems that premium users don’t typically raise, which helps optimize their models.
Finally, ThinkDeeper is also a response to growing competition. With Google integrating Gemini into Workspace, DeepSeek disrupting the market, and OpenAI preparing the response, Microsoft is looking to strengthen the loyalty of its user base.
Limitations and controversies: What Microsoft doesn’t say
Despite its advantages, ThinkDeeper faces several challenges. Each query consumes up to ten times more resources than a standard response, justifying the strict message limits and reduced speed.
Additionally, the tool does not reveal where it pulls its information from, unlike the standard Copilot, raising concerns about transparency.
Another worrying aspect is the possible algorithmic bias, since ThinkDeeper depends on data until October 2023, ignoring recent advances in fields such as medicine or legislation.
Some experts also warn that overuse of this tool could discourage critical thinking, leading to over-reliance on automated responses.
The future: Towards an “always deep” AI?
ThinkDeeper is just the beginning. Internal sources suggest that Microsoft plans to integrate it into applications such as Excel to optimize spreadsheets and Word to improve academic essays.
The launch of O3-mini and specialized versions of ThinkDeeper for engineers, doctors or educators are also expected.
More than a function, a declaration of intent
Opening ThinkDeeper to free users redefines the value of accessible AI. By combining the power of OpenAI with its global ecosystem, Microsoft not only competes with Google or Meta, but also raises a disturbing question: Are we prepared to live with tools that, in certain areas, reason better than us?
ThinkDeeper is part of a future where AI not only responds, but thinks with us. And although today it has limits, its mere existence forces us to rethink how we learn, work and solve problems.
Have you tried ThinkDeeper yet? Tell us your experience. Does it live up to the hype? Or is it still green? The discussion, like AI, is more alive than ever.
This post is also available in: