OpenAI Launches First Open Source Language Model, Competes With Meta, DeepSeek

At a press conference, OpenAI co-founder and president Greg Brockman announced two AI models, gpt-oss-120b and gpt-oss-20b, and emphasized that this was the first time the company had "released" an open-source language model in a long time. These new models only process text, are high-performance, and low-cost.

According to OpenAI, gpt-oss-120b and gpt-oss-20b are designed to run easily on local computer systems and are said to be suitable for AI tasks such as searching the Internet or running computer programs. The open source models are also tuned to prevent malicious use.

The tech company's leadership hopes the newly released model will aid in conducting new types of research and creating new products.

OpenAI also said it is working with partners, including French telecommunications giant Orange and US cloud-based data platform Snowflake, to put the new model into practice.

The move comes as the creator of ChatGPT is under pressure to share open source models. US-based Meta and Chinese-based DeepSeek are developing AI in this direction.

Among them, DeepSeek Company shocked the technology industry with its low-cost and high-performance AI model, which is completely open-source and community-oriented.

The term open source appeared since the birth of the software field, in which the source code of a system is publicly released on shared platforms, widely accessible, anyone can research, modify and develop applications on it.