Portuguese 
English 
Spanish 
3 min of reading 
0 comments 
New AIs Can Be Customized by Users and Companies and Are Already Available on Platforms Like Hugging Face and Amazon Bedrock
The OpenAI Launches Open AI Models Capable of Simulating Human Reasoning, Marking a Turnaround in the Company’s Stance on Transparency and Collaboration. The Announcement Was Made This Tuesday (5) and Breaks a Six-Year Hiatus Without Any “Open” Model from the Organization.
The GPT-oss-120b and GPT-oss-20b models are now accessible for free on the Hugging Face platform. With Advanced Capabilities Such as Text Generation, Programming, and Active Internet Search, the Systems Have Open Weights, Allowing for Customization by Developers, Companies, and Even Governments — But They Are Not Fully Open Source, as the Training Data Remains Protected.
What Is Behind OpenAI’s Open Models?
According to OpenAI Itself, the New Models Have Robust Performance but Focus on Efficiency: the GPT-oss-120b Runs on a Single 80 GB GPU, While the 20b Can Even Run on Laptops with 16 GB of RAM. This Allows Users Without Supercomputing Infrastructure to Experiment and Fine-tune Models Directly on Their Devices.
-
In a flooded well, archaeologists found a 1,700-year-old Roman egg that still contains clear and yolk intact inside the very thin shell.
-
Something is happening around the Earth: Inside the huge explosion of fireballs in 2026
-
A hot air bubble coming from Argentina expands over Brazil, causing thermometers to exceed 38 degrees with a thermal sensation of 40 degrees in late March, affecting 6 states at once.
-
The radish leaf that almost everyone throws away has more polyphenols, flavonoids, and fiber than the consumed root, and a 2025 study showed that the leaf contains compounds that protect the intestine, combat inflammation, and may inhibit the growth of cancer cells.
Despite Being “Open,” the Systems Do Not Have the Source Code Released. The Company Chose to Publish Only the “Weights” — the Numerical Parameters Adjusted During Training. This Decision Was Described by Sam Altman, OpenAI’s CEO, as a “Initial Step” Towards Building a New Open Source Model Following Pressure from Competitors Like DeepSeek from China, Which Gained Attention for Releasing Its R1 Model in Early 2025.
Why Did OpenAI Change Its Strategy Now?
The Last Time OpenAI Released an Open Model Was in 2019 with GPT-2. Since Then, the Company Adopted a Closed and Commercial Model, Focusing on Paid Products Like ChatGPT. The Recent Change Comes in the Wake of the AI Action Plan of the Donald Trump Administration, Which Encourages Open Models as Pillars of Digital Sovereignty and Global Standards.
The Change is Also Strategic. With the Advancement of Competing Companies in Customizable AI Systems, OpenAI Seeks to Regain Ground by Offering Controllable Alternatives for Companies That Wish to Run AIs on Their Own Servers — a Critical Point for Sensitive Sectors Like Military, Financial, and Governmental.
What Is the Impact on the Market and Research?
The Models Will Be Distributed Across Various Platforms: In Addition to Hugging Face, They Will Also Be on Amazon Bedrock, the Groq Infrastructure, and on the Servers of Humain AI from Saudi Arabia. According to OpenAI, Companies Like Orange and Snowflake Are Already Testing the New Systems.
During a Press Conference, OpenAI Reinforced That It Is Gathering Feedback to Define the Next Steps. Sam Altman Was Direct: “We Trust That the Community Will Create Amazing Things, but Once the Weights Are Released, There Is No Turning Back.”
Even So, the Company Postponed the Launch for Weeks to Conduct Security Tests and Risk Review, in Response to Criticism Pointing Out Open Models as Potential Vectors for Misinformation, Cybercrime, and Ethical Violations.
And You? Do You Believe That Open Models Are the Future of AI — or Do They Represent a Hard-to-Control Risk? Leave Your Opinion in the Comments. We Want to Hear Your Views on This Turnaround from OpenAI.
Seja o primeiro a reagir!