OpenAI Launches 2 GPT-OSS Models—Runs on Phones & Laptops
OpenAI launches GPT-OSS open models—120B & 20B—runs on laptops & phones. Apache 2.0, fine-tunable, MoE-powered.
"AI Disruption" Publication 7300 Subscriptions 20% Discount Offer Link.
Five years after their last open-source release, OpenAI has officially unveiled two open-weight language models: gpt-oss-120b and gpt-oss-20b. The last time OpenAI open-sourced a language model was back in 2019 with GPT-2.
OpenAI is truly embracing openness this time.
The AI community was buzzing with excitement today as OpenAI open-sourced gpt-oss, Anthropic launched Claude Opus 4.1 (detailed coverage below), and Google DeepMind released Genie 3. The three tech giants coincidentally dropped major bombshells on the same day, staging a clash of titans.
OpenAI CEO Sam Altman expressed uncontainable excitement on social media: “gpt-oss is out! We’ve built an open model that matches o4-mini performance and can run on high-end laptops. Super proud of the team—this is a major technical win.”
Model Highlights:
gpt-oss-120b: A large-scale open model designed for production, general-purpose, and high-inference use cases. It runs on a single H100 GPU (117 billion parameters, 5.1 billion active parameters) and is suitable for data centers, high-end desktops, and laptops.
gpt-oss-20b: A mid-sized open model optimized for low-latency, local, or specialized use cases (21 billion parameters, 3.6 billion active parameters), capable of running on most modern desktops and laptops.