Verified, then logged.
Sun 22 Nov 2026 · 00:23 KST / 15:23 UTC한국어 | ENSearch /

01 / AI · Analysis

AMD brings AI agent inference to neural chips built into PCs

Unsloth Desktop lets AI models run inference on a neural processing unit built into a PC.

AMD announced support for local inference across more than 100 AI agent apps and workflows on PCs with compatible built-in neural processing units (NPUs), according to Korean AI news outlet AI Times.

The report said AMD aimed to expand the use of AI agents that run on PCs without relying on cloud servers.

AMD brought that support to Unsloth Desktop, a platform for running AI models, according to the report.

How inference runs on the PC

AI Times reported that AMD had integrated its XDNA 2-based NPU with Unsloth Desktop.

In the setup described by the report, the calculations an AI model performs to produce a result take place on the PC itself.

What the support count covers

According to AI Times, the integration marked the first time Unsloth Desktop supported an NPU.

The reported figure counts the apps and workflows that can use local inference. The report did not give inference speeds for individual apps.