OpenAI Releases Open-Weight AI Models After 5 Years: What’s the Big Deal?
6 Aug, 2025 13:20 IST|Sakshi Post
For the first time in over five years, OpenAI has released two open-weight AI reasoning models, marking a significant shift in its approach to technology sharing. This move comes amid growing pressure from the rise of China’s open-source AI ecosystem, particularly through the emergence of DeepSeek, and criticism that OpenAI has drifted from its original mission of building openly accessible AI tools.
The two models — gpt-oss-120b and gpt-oss-20b — are now freely available for download on Hugging Face. Designed to be more accessible, they don’t require high-end hardware.
While gpt-oss-120b, the more powerful of the two, can run on a single Nvidia GPU, gpt-oss-20b, the lighter version, can even run on a consumer laptop with 16GB of RAM.
This marks OpenAI’s first open model release since GPT-2 back in 2019.
Open-Weight vs. Open-Source: What’s the Difference?
OpenAI’s new models are open-weight and not open source, meaning they include only the trained model weights. However, they do not come with the full source code, training data, or detailed model architecture. That means users can run the models but cannot fully inspect, customize, or retrain them.
By contrast, open-source models provide full transparency into the codebase, access to training algorithms and architecture, permission to modify, reuse, and distribute freely under open licenses, and ideally provide access to training data — although legal restrictions can limit this
Why Now?
The release reflects OpenAI’s response to global AI competition and political pressure. China’s DeepSeek, a powerful open-source model, has raised the bar for transparency and accessibility, pushing China to the forefront of open-source AI innovation. This shift has caught the attention of the U.S. government, which has been encouraging American tech companies to be more open in order to stay competitive.
Even OpenAI CEO Sam Altman recently admitted the company may have been "on the wrong side of history" regarding openness in AI development.
