**OpenAI Launches New Open-Weight AI Models: gpt-oss-20b and gpt-oss-120b**
OpenAI has announced the launch of two fresh open-weight AI models, gpt-oss-20b and gpt-oss-120b, representing a major advancement in the availability of AI technology. These models are now free for download, fulfilling OpenAI’s earlier pledge to release open-weight models this year.
### What Are Open-Weight Models?
Open-weight models are AI frameworks that can be downloaded and run on local devices, removing the necessity for an internet connection to access a model supplier’s server. This capability allows developers to craft tailored applications utilizing these models, promoting innovation and experimentation in AI creation.
### Model Specifications
OpenAI designates the gpt-oss-20b as a medium-sized model, whereas the gpt-oss-120b is regarded as a large-sized model. The smaller version is geared towards systems equipped with at least 16GB of VRAM or unified memory, making it ideal for high-end consumer GPUs and Apple Silicon Macs. Conversely, the larger model necessitates a minimum of 60GB of VRAM or unified memory, which makes it more resource-intensive in terms of hardware requirements.
### Demand for Open-Weight Models
The launch of these models is a reaction to the climbing pressure from competitors, especially after the emergence of DeepSeek in China. OpenAI had not unveiled an open-weight model since 2019, and these new offerings aim to address the increasing demand for accessible AI resources from the team behind ChatGPT.
### Real-World Performance
Although the gpt-oss models may not represent the most powerful AI solutions available from OpenAI, they are intended to provide solid real-world performance. Sam Altman, CEO of OpenAI, underscored the potential of the gpt-oss models, spotlighting their practicality and effectiveness across a range of applications.
### Conclusion
OpenAI’s introduction of the gpt-oss-20b and gpt-oss-120b models signifies a notable move towards making advanced AI technology more accessible. Developers and enthusiasts can now utilize these open-weight models to devise innovative solutions without the limitations of server dependence. For further details on these models, please visit OpenAI’s official website.
