Over the years, Microsoft-backed OpenAI has developed its fair share of language model systems, including GPT-4, which is a large multimodal model with text and image inputs support, DALL·E (AI to generate and edit images), Whisper (audio-to-text), Embeddings, Moderation, and more.
According to a new US trademark application submitted on July 18, Microsoft-backed OpenAI may be planning to launch another large language model, ‘GPT-5’. Spotted by us over the weekend, OpenAI has filed a new trademark application with USPTO for “GPT-5”, a “downloadable computer software for using language model”.
OpenAI has previously used the same “downloadable computer software for using language model” description in the trademark filing of previous generation models, such as GPT-4 and GPT-3.5. Unfortunately, the “GPT-5” name is the only interesting detail the listing reveals, and it doesn’t mean OpenAI plans to launch the new version this year.
We’ve done some digging into what OpenAI’s GPT-5 could be. In the filing, OpenAI mentions everything from “downloadable computer programs and software for using language models” to software for “the artificial production of human speech and text.”
It highlights possible features like “natural language processing, generation, understanding, and analysis.”
Additional functions include machine-learning-based language and speech processing, text and speech translation from one language to another, and software for machine…