Open AI finally launched GPT OSS after 5 year: A New Era of Open Weight AI Models

Open AI officially entered the open - weight with the release of its ground breaking chat GPT OSS models. Open AI announced on 5 August 2025 and this marks Open AI's first open weight large language model release since GPT 2 in 2019. With two powerful models GPT OSS 120b and GPT OSS 20b. The open AI company aims to offer high performance language capabilities to developers and research and business without restrictions on uses. This bold move reflects an OpenAI evolving strategy to balance innovation in the rapidly growing AI space.

What is GPT-OSS ?

GPT OSS stands for “open source series” though technically these models are open weight and not open source in the full traditional sense. The GPT OSS 120 b model features 117 billion parameters and uses a mixture of expert architecture and each of the 36 layers in the model contains 128 experts but only 4 experts are active for tokenization which makes it both powerful and computationally efficient.

In simple language, this architecture allows the model to selectively activate the most relevant part of its network for any given input and improve efficiency and scalability. This has to maintain performance while reducing computer cost and is an important factor in deploying such a large model at scale.

On the other hand, GPT OSS 20b is designed for lighter applications and with around 21 billion parameters. It’s only target devices with 16GB VRAM or higher which makes it a practical choice for developers building AI powered features into consumer level apps and tools

Key Features

Apache 2.0 Licence - The most exciting aspect is that both models are released under the Apache 2.0 licence and this means they can be used commercially modified and distributed opening doors for a wide range of applications from startup to internship.

Competitive Performance - GPT OSS 120b matches or even exceeds the performance of open AI proprietary o4-mini model on many reasoning and coding tasks. Meanwhile, GPT OSS 20b rivals o3 mini, which makes it highly competitive in the open weight category.

Safety Measures - OpenAI has definitely focused on safety and both models underwent rigorous testing under the preparedness framework, including adversarial fine tuning and red teaming. External researchers were also involved to evaluate potential risk and guide responsible use.

Why is this important?

The release GPT OSS comes at a time when AI openness is becoming more critical and with increasing demand for transparency and customizability in Ai tools and mainly developers and companies are turning toward open weight model until now matter LLaMA and Mistral’s mixtral were leading the pack but now open AI entry significantly change the landscape. 

OpenAI not only promotes innovation but also ensures that smaller organisations can access top tier. AI capabilities could lead to surge in new AI applications from education to healthcare entertainment and beyond.

Conclusion 

Open AI launch of GPT OSS is more than just a model release, their claim is it signals the company's readiness to share its powerful tools with the border AI community while it is still prioritizing safety whether you are a developer, researcher or business leader. This offers a new and powerful way to integrate.