{"id":25330,"date":"2025-05-12T17:10:51","date_gmt":"2025-05-12T15:10:51","guid":{"rendered":"https:\/\/comunicagenia.com\/blog\/reasoning-models-in-ai-are-they-approaching-a-limit\/"},"modified":"2025-05-12T17:10:51","modified_gmt":"2025-05-12T15:10:51","slug":"reasoning-models-in-ai-are-they-approaching-a-limit","status":"publish","type":"post","link":"https:\/\/comunicagenia.com\/en\/blog\/reasoning-models-in-ai-are-they-approaching-a-limit\/","title":{"rendered":"Reasoning models in AI: Are they approaching a limit?"},"content":{"rendered":"

Imagine a world where machines not only calculate, but also reason in a way analogous to a human being: solving complex equations, proving theorems or generating flawless code instantly.\u00a0<\/span><\/p>\n

Today,<\/span> this dream is a reality thanks to AI reasoning models<\/b>, which have established themselves as essential tools in fields as diverse as mathematical research or advanced programming.<\/span><\/p>\n

However, a recent report from Epoch AI, a non-profit organization dedicated to researching trends in artificial intelligence (AI), warns that <\/span>the exponential pace of progress could stall sooner than expected<\/b>.\u00a0<\/span><\/p>\n

While some experts envision a next computational and economic \u201cceiling,\u201d others, like Sam Altman of OpenAI, continue to bet on new innovations. Are we on the verge of a slowdown or are we simply facing a temporary blip?<\/span><\/p>\n

Keep reading and <\/span>discover what could stop these advances<\/b>, the promises that still maintain optimism and the avenues of research that aim to overcome current barriers.<\/span><\/p>\n

Why might progress be slowed?<\/span><\/h2>\n

Epoch AI maintains that the escalation in resources needed to train reasoning models is reaching unsustainable levels.\u00a0<\/span><\/p>\n

For example, <\/span>to move from model o1 to o3, OpenAI invested approximately ten times more computing power<\/b>, concentrated in refinement phases through reinforcement learning.\u00a0<\/span><\/p>\n

If this trajectory continues, it is estimated that the development of the following generations could exceed 1 billion euros in 2027, a figure that puts the margins of any research laboratory in check.<\/span><\/p>\n

Infrastructure limits<\/span><\/h3>\n

The production of semiconductors and the availability of electrical energy are two critical bottlenecks. <\/span>Data centers already consume a high percentage of global energy<\/b>, and the manufacturing of <\/span>advanced chips<\/span><\/a> faces supply constraints and rising costs.\u00a0<\/span><\/p>\n

According to Epoch AI, up to 2\u00b710\u00b2\u2079 FLOP would be needed to achieve the 2030 goals, a magnitude that far exceeds the current installed capacity today.<\/span><\/p>\n

Diminishing returns<\/span><\/h3>\n

Traditional \u201cscaling laws\u201d indicate that each increase in power or parameters produces improvements in performance that follow a curve of diminishing returns.\u00a0<\/span><\/p>\n

That is, as the model and the computation grow, the gain per extra unit of resource decreases, <\/span>making each advance more costly and less profitable.<\/b><\/p>\n

Reasoning models have relied on scaling both size and computation (and also inference time) to break records in specialized benchmarks. Epoch AI and other analyzes project that it will soon no longer pay to scale just by adding computing power.<\/span><\/p>\n

What do reasoning models promise us?<\/span><\/h2>\n

In the previous part we talked about <\/span>why<\/b> we could reach a brake (costs, infrastructure, diminishing returns). So it’s time to remember why we invest so much in these models.<\/span><\/p>\n

Solutions to complex problems<\/span><\/h3>\n

Despite the shadows, the progress has been surprising. Models like o3 or DeepSeek-R1 can:<\/span><\/p>\n