# Nueva arquitectura de IA de startup ofrece razonamiento 100 veces más rápido que los LLMs

**URL:** <https://www.yodev.dev/t/startup-drops-new-ai-architecture-that-delivers-100x-faster-reasoning-than-llms/204>\
**Category:** AI Dev Tools — General\
**Tags:** ai-data-sci\
**Created:** [28 Julio, 2025 16:10 UTC](https://www.yodev.dev/t/startup-drops-new-ai-architecture-that-delivers-100x-faster-reasoning-than-llms/204 "2025-07-28T16:10:59Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Devy](https://yyz1.discourse-cdn.com/flex009/user_avatar/www.yodev.dev/devy/32/62_2.png) [@Devy](https://www.yodev.dev/u/Devy)\
**Post date:** [28 Julio, 2025 16:10 UTC](https://www.yodev.dev/t/startup-drops-new-ai-architecture-that-delivers-100x-faster-reasoning-than-llms/204/1 "2025-07-28T16:10:59Z")

</div>

**FROM THE FRONTIER - via SuperHuman**

 ![](https://canada1.discourse-cdn.com/flex009/uploads/inovacon/original/1X/6da9b74e93d0be6210402c1b6948bc16eb7e13aa.webp)

LLMs have a big problem. Unlike humans, who can reason internally without verbalizing every thought, LLMs are forced to express all reasoning as text (aka chain-of-thought prompting). This approach is inefficient — it requires copious amounts of training data and can fail if steps are slightly out of order. The result: AI that’s data-hungry, computationally expensive, and still falls short on complex reasoning.

A startup from Singapore set out to fix this bottleneck. Sapient Intelligence has unveiled the [Hierarchical Reasoning Model](https://link.mail.beehiiv.com/ss/c/u001.mzLE_NAgQL1lHxwbP_X58Pi4HHEaRYUBg6hF6zqszPlbOnmH6HbyCe3UGIxpbi3klrAKLzkZAxHIUSjFuHEjJAZh8EAmTr-ZN3epokzIGu44-gur1UNSkp6b07hXpIMOohz3kEwnwNzxsopJwJCJ4XXVOjPEiGhSyxQ6qo0m1s-4Ubh3VS4nUvKlEKcoCOpX_surFbPWgkKwtLcPx85ozJ9YXvKYnewKqgj3tdSwHA8R_CK3hwXds7dJe95svqeg/4ik/AMl8wmZnTf2iYZ9vUj_hig/h12/h001.5otw4iuDf43H0ry7J6wsmkrDoEQy2LysrS1uXAc1F8s)(HRM), an AI architecture that beats traditional LLMs on complex reasoning tasks with a fraction of the data and compute.

 ![](https://canada1.discourse-cdn.com/flex009/uploads/inovacon/original/1X/7cf594ab71cf01ea8ae4c85d85992b89fffca447.webp)

_Sapient’s HRM architecture outperforms traditional models on complex reasoning tasks. Source: arXiv_

Here’s how it works: HRM [mimics](https://link.mail.beehiiv.com/ss/c/u001.L3HibA3BiJ236dOrVqQGNvwjlC_O7NRz8AVnD5g5UciJZ78iN0R0PY-ft4E7vaGwLsOV8TPMP9UUXFwcfzfRV-VgbCOnSMZ1C-JdM8Be5XwcbM-IA36AvwwT2A0TeXMwSMTmNTv-Veftxuzdqgi8qIDcII3XUcs7NL2YwPbsN_4TxJgXyGwwP5urFEyetT7xV2u42CJIDvsc2-vMLDMWW3unVpgDEq7PyofXsSDz8_ULLaaQSIlI-CjOfkIFxXsEUhAL1iBvwkNaJg_KHUNiHUrsQs_vl1WDn5RyPSviAJLtvg2C_OP96STOjRj1oOdb2rkjzkfEbMwrTekT6dAa_a6RmSb8QAxTmlW8qcD2tsYpN0pJO34UkHleM-jrWWqN/4ik/AMl8wmZnTf2iYZ9vUj_hig/h13/h001.4jUkO4yxF9_sNwBh7jAylJItHYLJcgFC1PNW_gp5Mvs) how human brains think. It uses 2 interconnected modules that work together like a supervisor and a worker. The worker module crunches the numbers on a sub-task until it reaches a solution. Then, the supervisor evaluates this progress, updates the overall strategy, and assigns the worker a new sub-task. This back-and-forth lets the AI think without verbalizing each step as text, requiring a fraction of the compute that goes into traditional models.

This could have significant implications:

- **David beats Goliath:** HRM blew much larger AI models out of the water on complex reasoning tasks with 100x faster processing, achieving near-perfect scores on challenges where state-of-the-art models completely failed.

- **Real-world efficiency:** For enterprises, this translates to major cost savings and new capabilities, with training times slashed to hours instead of months.

- **New frontiers:** This approach shows promise in fields that require complex decision-making, like healthcare, climate forecasting, and robotics.