GPT-5 mini (medium) is a multimodal language model developed by OpenAI as part of the GPT-5 family. Released in August 2025, it is positioned as a mid-tier reasoning model that balances high-level intelligence with operational efficiency. The model is categorized within the medium-sized class, estimated to contain between 40 billion and 150 billion parameters, and is designed to handle complex tasks requiring deep reasoning and synthesis.
Key capabilities of the model include support for both text and image inputs with a large context window of 400,000 tokens. It is optimized for agentic workflows and multi-step reasoning, though benchmarks indicate it operates at a slower generation speed (approximately 88 tokens per second) compared to non-reasoning variants in the same class. The model's knowledge cutoff is cited as May 2024.
In the OpenAI ecosystem, GPT-5 mini (medium) serves as a bridge between the highly specialized flagship models and the more lightweight "nano" variants. While it offers significant improvements in reasoning over previous generations, its performance comes at a premium pricing tier for output tokens, reflecting its focus on high-precision tasks and logical consistency.