Author: IBL News

• “The gpt-oss-120b model achieves near-parity with OpenAI o4-mini on core reasoning benchmarks, while running efficiently on a single 80 GB GPU.”

• “The gpt-oss-20b model delivers similar results to OpenAI o3‑mini on common benchmarks and can run on edge devices with just 16 GB of memory, making it ideal for on-device use cases, local inference, or rapid iteration without costly infrastructure.”

• “These models are compatible with our Responses API⁠ and are designed to be used within agentic workflows with exceptional instruction following, tool use like web search or Python code execution, and reasoning capabilities.”

OpenAI trained the models on a mostly English, text-only dataset, with a focus on STEM, coding, and general knowledge.

Therefore, these new text-only models are not multimodal, but they can browse the web, call cloud-based models to help with tasks, execute code, and navigate software as an AI agent. The smaller of the two models, gpt-oss-20b, is compact enough to run locally on a consumer device with more than 16 GB of memory.

OpenAI defined its release as “best-in-class open models”, highlighting that they work anywhere: locally, on-device, or through third-party inference providers. The company partnered with several deployment platforms such as Azure, Hugging Face, vLLM, Ollama, llama.cpp, LM Studio, AWS, Fireworks, Together AI, Baseten, Databricks, Vercel, Cloudflare, and OpenRouter.

The fact that the “weights” are publicly available means that any developer can peek at the internal parameters to get an idea of how it processes information. They can work as a complement to OpenAI’s paid services. Unlike ChatGPT, you can run a gpt-oss model without an internet connection and behind a firewall.

With Apache 2.0, models can be used for commercial purposes, redistributed, and included as part of other licensed software.

Open-weight model releases from Alibaba’s Qwen as well as Mistral, also operate under Apache 2.0.

To try the models, OpenAI released an open model playground⁠, alongside guides.

In the U.S., the open-weight leader has been Meta. The tech giant released the first of its Llama series of models back in 2023, with Meta’s most recent release, Llama 4, arriving a few months ago.

Also, the Chinese startup DeepSeek released its cheap-to-run model that was open-weight this year.

• Microsoft Azure: OpenAI’s open‑source model: gpt‑oss on Azure AI Foundry and Windows AI Foundry

• NVIDIA: OpenAI’s New Open Models Accelerated Locally on NVIDIA GeForce RTX and RTX PRO GPUs