OpenAI’s Ultrafast And Decisions API Shift AI Race To Speed, Cost

Frontier model companies are leaning more heavily into high-powered models that are faster and cheaper, but sometimes trading one for the other. OpenAI’s latest push to make AI faster and more economical are two new offerings that address different parts of that AI performance equation. The company is rolling out Ultrafast for GPT-6.1 Sol, promising up to eight times faster token generation in Codex, alongside a new Decisions API that handles the quick judgments AI applications and autonomous agents make throughout their work.

The announcements reflect a growing industry focus on how much time and money AI consumes to get useful work done, more so than simple performance on benchmarks.

The question of cost and speed is becoming increasingly important as companies move from prompt and chat-based interactions to longer-lived automated workflows involving hundreds or thousands of model calls.

Saving a few cents or milliseconds on an individual request might seem insignificant, but those savings add up when businesses process millions of transactions. The economics become even more compelling when developers can use inexpensive specialized models for tasks that don’t require sophisticated reasoning.

Another factor driving the demand for faster, cheaper model access are open weight and foreign models that are steadily gaining market and developer share. Chinese AI models, including DeepSeek, Moonshot, Kimi and Alibaba’s Qwen, are competing aggressively on price and performance. Open-weight models provide alternatives to expensive proprietary APIs.

OpenAI Is Offering More Speed, But At A Price

OpenAI recently introduced GPT-6.1 Sol provides what OpenAI describes as near-Astra intelligence for agentic coding, computer use and professional tasks at approximately one-fifth of GPT-6 Astra’s standard token prices.

AI Doesn’t Need To Think So Hard About Everything

That’s an expensive way to answer yes or no or classify things into buckets.