OpenAI released two open-source language models, gpt-oss-20B and gpt-oss-120B, marking its first open-weight model launch since GPT-2 in 2019 and its first open-source foundation models since CLIP in January 2021. This is a huge, surprising shift that reverses OpenAI’s recent trend of proprietary releases and restores its commitment to transparency and community-driven innovation.
This announcement represents a major strategic pivot for OpenAI, responding to developer demand for accessible model weights and mounting competition from open-weight offerings by DeepSeek and Meta’s LLaMA. By open-sourcing these high-capacity reasoning models, OpenAI allows researchers and enterprises to run inference locally, while also offering transparency and customization.
And such was the impact in the community that gpt-oss-120B broke into the top 50 most liked models of all time in a time period of less than 24 hours, surpassing the total like count of giants such as DeepSeek-R1-0528, grok-1, Llama-2-70b-chat-hf, and even other OpenAI's open source models like whisper-large-v2 and clip-vit-large-patch14.
Thus, gpt-oss-120B was comfortably sitting at position #43 among the most liked models of all time just one day after its release. Similarly, gpt-oss-20B, the lighter version of the two, has also been performing well. Reaching position #61 after the 24 hour-mark, it has surpassed the lifetime likes of models like stable-diffusion-2, whisper-large-v2 and Qwen2.5-Omni-7B. Given this momentum, I predict it is only a matter of time before these models break into the top 25, and even the top 10, most liked models of all time.
You can learn more about these new models by reading our story and OpenAI's post.