The OpenAI intelligent framework took a strategic turn and has been a benchmark for innovation in artificial intelligence since its inception.
However, this past Tuesday has brought two news that, at first glance, seem contradictory: the unexpected delay of its first open weights model since GPT-2 and the launch of o3-pro, the premium version of its reasoning engine.
In this duality lies OpenAI’s great challenge: finding the balance between the transparency demanded by the sector and the rapid delivery of high-performance technological solutions.
Below, we explore both announcements and analyze how they fit into the company’s overall strategy.
The delay of the open model: a step back for transparency?
When a leading company announces that it will release the “weights” of one of its models, the community of researchers and developers expects a real turning point.
Open weights allow you to study, modify and build new applications on a base architecture, promoting transparency and accelerating innovation thanks to external contributions.
OpenAI, often criticized for the closed nature of GPT-4, had announced its new open weights model earlier this summer. However, on June 10, 2025, it confirmed that its release would be postponed until the end of the season.
According to Sam Altman, the delay is not due to technical problems, but to an “unexpected and astonishing” finding during the final validation phase.
With this distinction, OpenAI suggests that sacrificing timing is worth it to incorporate substantial improvements. However, the lack of details has unleashed a wave of speculation in the community.
Impact on perception of openness
By postponing the launch, OpenAI shows how difficult it is to meet a strict schedule without giving in to the unforeseen events of cutting-edge research.
But this delay also opens up a competitive edge. The company now competes against already established open source projects, such as DeepSeek R1 or the recently introduced Magistral family of models from Mistral.
In this context, the community observes with some skepticism whether the OpenAI proposal will arrive on time and without the restrictions that usually accompany commercial products.
o3-pro: the commitment to the professional elite
On June 10, OpenAI introduced o3-pro, the premium variant of its powerful o3 reasoning model. This model is aimed at Pro and Team users, with plans to extend it to Enterprise and education sector customers.
o3-pro debuts advanced features such as native web search integration, file analysis, visual understanding and Python code execution. Its API price ($20 per million tokens in and $80 per million tokens out) positions it in a professional segment where quality and rigor prevail over cost.
o3-pro’s focus on corporate and academic environments responds to a clear demand: applying high-level AI in areas such as scientific research, legal assistance or financial analysis.
As expected, its high precision and ability to follow complex instructions mean higher latency than previous models.
Open vs. open innovation closed ecosystem: the great contradiction
The launch of o3-pro consolidates OpenAI’s role as a provider of closed, top-level solutions, disconnected from the pace of the open source community.
While the open model takes its time to be perfected, o3-pro expands the catalog of proprietary tools that lead cutting-edge technology.
The contradiction is evident: on the one hand, a gesture of transparency with the promise of an open model; on the other, the creation of new economic barriers that limit access to their most advanced capabilities.
A two-sided strategy
In June 2025, OpenAI travels down a path of complex decisions. The academic community and startups demand the freedom to innovate without depending on expensive licenses or APIs.
By delaying its open model, OpenAI risks losing ground to competitors that already offer robust solutions. At the same time, with o3-pro, it reinforces its ecosystem of paid products that generate million-dollar income to finance its research.
This duality, far from being a simple contradiction, could be a calculated strategy: use the income from its exclusive products to sustain long-term research projects, the fruits of which could eventually benefit everyone.
The big question is whether this balance will satisfy both audiences or whether it will end up further polarizing the community between those who advocate open and collaborative development, and those who can afford access to the AI elite through commercial licenses.
The future of OpenAI
These two announcements encapsulate the duality that defines OpenAI today: a company that pursues technical excellence while navigating a fiercely competitive environment.
The delay of their open model, while frustrating to some, reflects a commitment to quality, suggesting that they are not willing to launch a half-baked product. On the other hand, o3-pro demonstrates its agility to meet the demands of an increasingly demanding professional market.
The real test will come when your open model sees the light. That’s when it will be known if OpenAI can balance its promises to the community and its dominance in the market.
This post is also available in: