OpenAI has launched two open AI models with human-like reasoning capabilities. This move follows the success of DeepSeek, a Chinese AI company. The models are freely available.
The models, named gpt-oss-120B and gpt-oss-20B, are OpenAI's most capable open-source models to date. They outperform similarly sized models in reasoning tasks and are optimised for efficient deployment. The models also demonstrate strong tool use capabilities. OpenAI trained the models using reinforcement learning techniques informed by its most advanced internal models.
OpenAI's decision to release open-source models marks a shift since 2019 and positions it against competitors like Meta and Mistral. These models allow developers to build upon them to improve their own products. Sam Altman, CEO of OpenAI, expressed excitement about making the models available to the world, hoping to enable new research and product creation.
