OpenAI is taking a bold step into the open-source world. The company plans to release its first open AI model since GPT-2 — and it’s promising to set a new benchmark for reasoning capabilities in the AI community.

Announced in March and now beginning to take shape, the model is reportedly being developed under the leadership of Aidan Clark, OpenAI’s VP of Research. Sources close to the company say OpenAI is targeting an early summer 2025 release. The goal? Deliver a model that tops all other open reasoning models — and does so under a highly permissive license.

A Shift in Strategy

This move marks a significant departure from OpenAI’s historically closed approach. CEO Sam Altman has publicly acknowledged the need to rethink the company’s open-source strategy, stating in a Reddit Q&A, “I personally think we need to figure out a different open source strategy.”

While not all OpenAI leadership agrees with this direction, the momentum appears strong — especially with competitors like Meta’s Llama and China’s DeepSeek gaining traction by openly releasing their own models. Meta has reported over 1 billion downloads for its Llama models, and DeepSeek is quickly building a loyal global user base.

What to Expect from the Open Model

OpenAI’s upcoming release will be a “text in, text out” model, built for accessibility on high-end consumer hardware. It’s expected to include a toggleable reasoning mode — an option increasingly popular among frontier models like those from Anthropic. This allows developers to strike a balance between accuracy and latency, enabling more flexible real-world applications.

OpenAI is also emphasizing safety and transparency this time around. According to insiders, the model will be red-teamed extensively and released alongside a detailed model card, outlining internal benchmarks, safety tests, and usage guidelines.

“We will evaluate this model according to our preparedness framework, like we would for any other model,” said Altman on X last month. “And we will do extra work given that we know this model will be modified post-release.”

Pressure from the Open AI Ecosystem

OpenAI’s decision appears to be a direct response to increasing pressure from the broader AI ecosystem. Rivals have been rapidly releasing open models with more permissive licenses — something OpenAI’s past efforts have lacked. Google’s Gemma, for instance, received criticism for licensing restrictions, and OpenAI is aiming to avoid the same fate.

While many details are still under wraps, early signals suggest OpenAI’s open model could become a powerful new benchmark in the open-source AI space — and may be followed by smaller or more specialized variants depending on community reception.

Final Take

OpenAI is positioning its first open model not just as a token gesture to the community, but as a serious competitor in the rapidly evolving landscape of open-source AI. If successful, it could reset expectations for what “open” really means — and challenge the dominance of existing players like Llama, DeepSeek, and Mistral.

With a permissive license, high-end capabilities, and red-teamed safety baked in, OpenAI’s open model could become one of the most impactful releases of 2025 — and a signal that even the most closed labs can learn to open up.

Get the Latest AI News on AI Content Minds Blog