OpenAI shook up the AI scene on Tuesday by releasing two open-weight language models, making powerful reasoning tools free and downloadable. This move levels the playing field for developers and researchers who want to run advanced AI locally or behind their firewalls. 🚀
Unlike open-source models, these open-weight models give you access to trained parameters without revealing the entire codebase or training data. “One of the things that is unique about open models is that people can run them locally. People can run them behind their own firewall, on their own infrastructure,” said Greg Brockman, OpenAI’s co-founder, in a press briefing.
The AI landscape has been fierce this year. Meta’s Llama models once led the pack, until DeepSeek from the Chinese mainland released a powerful, cost-effective reasoning model. Now, OpenAI is back in the open-model game with its first open release since GPT-2 in 2019.
The two new models—gpt-oss-120b, which runs on a single GPU, and gpt-oss-20b, small enough for your personal computer—match the performance of proprietary reasoning models like o3-mini and o4-mini. They especially shine in coding 💻, competition math ➗, and health-related queries 🏥.
Trained on a text-only dataset focused on science, math, and programming, these models empower young innovators, entrepreneurs, and students to experiment and build without infrastructure barriers. Ready to download and customize, they’re a game-changer for anyone diving into the AI revolution.
Reference(s):
OpenAI releases free, downloadable models in competition catch-up
cgtn.com
