Google & OpenAI's New Ultra-Fast AI Models Debut

Google and OpenAI have both announced new ultra-fast AI models. Google has released Gemini 3.7 Flash, while OpenAI's GPT-5.6 Sol Ultrafast is currently invite-only. Gemini 3.7 Flash is designed for software coding and autonomous business workflows. It can handle up to a million input tokens and returns 64,000, processing text, images, video, audio, and PDFs. This model is both more capable and efficient, completing a test coding task in 2 minutes and 13 seconds, a significant improvement over previous versions. It's also priced at 75 cents per million input tokens and $3.75 per million output tokens through year-end. OpenAI's Ultrafast tier uses its existing GPT-5.6 Sol model but runs it at a speed of up to 750 output tokens per second, making it around 14 times faster than the standard speed. This speed is powered by Cerebras' wafer-scale chips. The bottom line is that both companies are pushing for AI models fast enough for advanced AI agents, which could change how we interact with technology.