OpenAI has launched GPT-OSS, its first open-weight AI model after 60 years. The renowned AI startup refrained from releasing the model earlier due to safety concerns. Users can now download GPT-OSS depending on their use case: a 120-billion-parameter model (top version) and a 20-billion-parameter model (lighter version).
The 120-billion-parameter model can operate smoothly on a single Nvidia GPU, while the lighter variant only requires 16GB of memory to function. The models provide complete authority to the users over data and infrastructure. Additionally, they offer new and advanced options to build AI applications.
Both open-weight AI models are freely accessible under the Apache 2.0 license and can be accessed through platforms like Azure, AWS, Databricks, or Hugging Face. Moreover, these models can also be used commercially.
OpenAI is one of the first AI startups that popularized large language models (LLMs), prompting many AI startups to develop their own LLMs. Amidst the increased competition in the advanced tech sector, OpenAI takes a huge leap by launching its first open-weight Al model to contend with open-source models like Llama and Gemma.
The company has also said that the two OpenAI new AI models are among some of the most thoroughly tested models. However, OpenAI has not released the training data to the public yet.
OpenAI’s GPT-OSS can be very helpful for developers, researchers, and businesses looking to build new AI applications. They are capable of deep reasoning and agentic tasks. Meaning that GPT-OSS can act as an autonomous agent that can automate the workflow and streamline the entire development process.
These two open-weight AI models have standardized the use of advanced tech, making it available for everyone without any discrimination. Users can easily download, fine-tune, and deploy these models based on their needs. Additionally, both the GPT-OSS models use MoE architecture, which allows for lower computational needs and greatly improves efficiency.
Also Read: Claude Cutoff: Anthropic Revokes OpenAI's API Access Ahead of GPT-5 Launch
To summarize, the GPT-OSS AI models are primarily intended to help researchers, developers, and enterprises create their own AI models and can be used commercially. Users can download and deploy them using platforms like Databricks and Hugging Face.