OpenAI launches free customizable AI models for developers

The move places OpenAI in direct competition with recent open-weight releases from Meta, Mistral AI, and China’s DeepSeek, as part of a growing trend toward transparency and developer autonomy in AI.

By  Storyboard18| Aug 6, 2025 12:27 PM
The move places OpenAI in direct competition with recent open-weight releases from Meta, Mistral AI, and China’s DeepSeek, as part of a growing trend toward transparency and developer autonomy in AI.

OpenAI has released two new open-weight AI language models, named gpt-oss-120b and gpt-oss-20b, marking its first open-weight release since GPT-2 in 2019. The announcement was made on August 5, with OpenAI positioning the models as low-cost, customizable alternatives for developers and researchers.

The text-only models are designed to run efficiently on different hardware, with gpt-oss-120b optimized for a single Nvidia GPU, and the lighter gpt-oss-20b capable of operating on consumer laptops with 16GB RAM. Both models are available for free via Hugging Face and GitHub under the Apache 2.0 license, and can also be used through tools like LM Studio and Ollama.

The move places OpenAI in direct competition with recent open-weight releases from Meta, Mistral AI, and China’s DeepSeek, as part of a growing trend toward transparency and developer autonomy in AI.

Open-weight models differ from open-source ones in that while they make their trained parameters public, they don’t necessarily release the full training code or data. This provides greater control and insight into how the models behave, without exposing proprietary architectures.

In a briefing, OpenAI noted the models can also route complex tasks to more powerful closed models in the cloud when needed, allowing for hybrid use cases.

Both models performed competitively in benchmark coding tests, scoring 2622 and 2516 on Codeforces, respectively, outperforming rivals like DeepSeek’s R1.

The release follows multiple delays. Last month, CEO Sam Altman said more time was needed to run safety tests and review high-risk areas before making the models publicly available.

The project includes hardware partnerships with Nvidia, AMD, Cerebras, and Groq, with Nvidia CEO Jensen Huang saying OpenAI is pushing “innovation in open-source software” built on Nvidia’s AI infrastructure.

First Published onAug 6, 2025 12:51 PM

OpenAI urges students to see ChatGPT as learning aid, not answer shortcut

Leah Belsky, the company's Vice President of Education, emphasized that AI literacy is now a foundational skill for graduates entering the modern job market.