← Retour au blog
tech 3 August 2026

Prioritizing Speed Over Intelligence in Model Selection: A Paradigm Shift

In a world where AI models reach sufficient intelligence for most daily tasks, speed becomes the primary selection criterion. Discover why and how this impacts tech businesses.

Article inspired by the original source
I'm (mostly) picking models on speed now, not intelligence ↗ martinalderson.com

Introduction

The landscape of artificial intelligence (AI) is constantly evolving, and a significant paradigm shift is occurring: model speed is beginning to outweigh raw intelligence. Martin Alderson discusses this phenomenon in an article where he explains why he now prioritizes speed over raw intelligence for his daily tasks. But what does this mean for tech businesses?

Why Speed is Taking Over

Historically, model intelligence has been the main selection criterion. However, with models like Opus 4.6 that are "smart enough" for common tasks such as coding, conducting research, or data analysis, speed becomes a critical factor. Studies show that users prefer faster software, even if it is less sophisticated.[1]

Indeed, the interaction speed with a model can significantly enhance the user experience. When a model is fast, it allows for a smoother and more efficient workflow, which is essential in increasingly time-pressured environments.

The 100tok/s Threshold

Alderson highlights an interesting threshold: 100 tokens per second (tok/s) is the equivalent of "100ms feeling instant" for humans. This is the speed at which the average user can keep up with the model without being overwhelmed. Below 50tok/s, the model seems slow, while above 200tok/s, it becomes almost unsettling.

This notion of speed is crucial as it directly influences user productivity. Fast models like GLM5.2 or DeepSeek V4 Flash GA, capable of exceeding 100tok/s, enable the completion of complex tasks more rapidly.

Impact on Tech Businesses

For decision-makers and entrepreneurs, this paradigm shift means it's time to reconsider the criteria for selecting AI models. Speed can lead to productivity gains and better user satisfaction, which is vital to remain competitive.

For example, a company using a fast model for data analysis can obtain results more quickly, speeding up decision-making processes. In another case, a fast model for customer support can significantly improve user experience by reducing waiting times.

Conclusion

As AI models continue to improve, it is essential to adapt our selection criteria to maximize benefits. Prioritizing speed might seem counter-intuitive at first, but it's a strategy that can pay off in the long run in a world increasingly focused on efficiency.

Let's discuss your project in 15 minutes.

AI models speed productivity technology decision-making
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call