OSS model added to OpenAI, let’s know about it..

OpenAI Just Went Open Source: Two Powerful GPT Models You Can Run Locally!

The Big Reveal

OpenAI has just released two open-weight AI models—gpt-oss-120b and gpt-oss-20b—marking its first public release of this kind since GPT-2 in 2019

What Makes These Models Stand Out

Both models are open-weight, meaning you get full access to their trained parameters under a permissive Apache 2.0 license, allowing commercial and personal use
gpt-oss-120b (117B parameters) delivers impressive reasoning and tool-performing capabilities—on par with OpenAI’s o4-mini
gpt-oss-20b (21B parameters) may be smaller but still packs a punch—and is comparable to o3-mini

Run it on Your Own Hardware

gpt-oss-20b can run on consumer-grade PCs with just 16 GB of RAM—think high-end laptops or desktops
gpt-oss-120b, while needing more muscle, can run on powerful setups like an 80 GB A100 GPU
These models support popular inference tools like vLLM, Ollama, llama.cpp, and are available via platforms like Hugging Face, AWS, Azure, and more

Why It Matters

This is a major shift toward transparency and democratization of AI:

Developers can now fully inspect and modify how the models reason and produce outputs
OpenAI has stressed privacy and control—these models run entirely on your own infrastructure, so your data stays yours
Safety hasn’t been overlooked—both models have undergone extensive checks, and OpenAI is even offering a $500,000 red-teaming prize to find potential vulnerabilities

Want to Get Started?

You can download the models from Hugging Face or visit the official OpenAI “Open Models” page for more info and tutorials
These models support fine-tuning, chain-of-thought reasoning, function calling, and more—so whether you’re a solo hobbyist or building the next-gen app, the possibilities are open-ended

OpenAI has unlocked two state-of-the-art GPT models—gpt-oss-120b and gpt-oss-20b—under open licenses. You can run them locally, inspect how they think, and build your own AI-powered tools without needing the cloud. A big win for developers and innovators everywhere.