OpenAI has launched two innovative artificial intelligence models with enhanced logical reasoning capabilities — gpt-oss-120b and gpt-oss-20b. These models represent the first open-weight offerings since the release of GPT-2 over five years ago. They are freely accessible on Hugging Face, geared towards developers and researchers looking to create their own solutions based on open technologies.
The models differ in power and hardware requirements:
- gpt-oss-120b — a more powerful model that can operate on a single NVIDIA GPU;
- gpt-oss-20b — a lightweight version that can run on a standard laptop with 16 GB of RAM.
OpenAI aims to present an open AI platform in the United States as an alternative to the growing influence of Chinese labs like DeepSeek, Qwen (Alibaba), and Moonshot AI, which are actively developing robust open models.
Regarding performance testing, the model scored 2622 points on Codeforces, with the 20b version scoring 2516, surpassing DeepSeek R1 but falling short of closed models like o3 and o4-mini. In the challenging Humanity’s Last Exam (HLE), the 120b achieved 19%, while the 20b reached 17.3%, outperforming other open models but trailing behind o3.
These new models have been trained using a methodology similar to that of OpenAI's closed models. They utilize a mixture-of-experts (MoE) approach, activating only a subset of parameters for each token, significantly enhancing efficiency. Additional reinforcement learning has enabled the models to develop logical reasoning chains and invoke tools such as web searching or executing Python code.
The models exclusively handle text and do not generate images or audio. They are distributed under the Apache 2.0 license, allowing commercial use without permission from OpenAI, although the training data remains proprietary due to copyright risks.
The launch of gpt-oss is intended to strengthen OpenAI’s position within the developer community and respond to political pressures in the U.S. aimed at enhancing the role of open American models in global competition.
