OpenAI has introduced gpt-oss-120b and gpt-oss-20b, two new open-weight reasoning models designed for advanced instruction following, tool use, and reasoning capabilities, under the Apache 2.0 license. These models incorporate a Mixture-of-Experts (MoE) design, allowing for efficient computation by activating a subset of parameters per token, and they support agentic workflows such as real-time web search and Python tool integration. Compared to other models like GLM-4.5, Qwen3 Thinking, DeepSeek R1, and Kimi K2, GPT-OSS demonstrates strong reasoning and math performance while maintaining a smaller active parameter footprint. Although GLM-4.5 excels in agentic workflows and function-calling due to its larger size, GPT-OSS offers a balance of performance and deployment efficiency, making it suitable for developers with limited computational resources. The release of these models underscores OpenAI's commitment to innovation and safety in AI, aiming to provide accessible, high-performance models for various applications.