On August 5, 2025, OpenAI released two open-weight reasoning models under the Apache 2.0 license, named gpt-oss-120b and gpt-oss-20b, allowing researchers to freely access, modify and distribute these AI models. This industry-milestone move responds to growing demand for open-source, high-performance models that make AI development more transparent. The two models utilise GPT-based architectures developed over the past two years and demonstrate particularly strong reasoning capabilities while making the models' weights and structure fully accessible.
The new models deliver outstanding performance, with the gpt-oss-120b model achieving an 87.8% score on the MATH benchmark, while the smaller gpt-oss-20b showed a 59.7% result, significantly outperforming open-weight models of similar size. During model development, OpenAI considered safety aspects and implemented built-in protection mechanisms against harmful use, as highlighted by Jason Wei, a member of OpenAI's research team. OpenAI made the models available on the HuggingFace platform and provided extensive documentation and sample code on GitHub to facilitate model usage across various hardware environments, including NVIDIA and AMD GPUs.
Experts suggest OpenAI's decision will significantly impact the AI sector, particularly in developing open-source models. The models demonstrated exceptional performance across 28 different tasks, including solving mathematical problems, coding, and reasoning capabilities. The Apache 2.0 license enables widespread use of the models for both commercial and research purposes, which is expected to accelerate AI innovation and make advanced AI technology more accessible to smaller organisations and developers.
Sources:
1.
openai.com
2.
Maxwell Zeff
3.
Servifyspheresolutions
