# Sam Witteveen — 'Dory - NVIDIA's Looped Transformer'

> Sam Witteveen walks through NVIDIA's Dory, a 4B model that spends more compute per token by looping its middle layers.

- Org: Sam Witteveen
- Date: 2026-11-02
- Category: video
- Source: https://www.youtube.com/watch?v=vSCYL8ySh9g
- Canonical page: https://ai-tldr.dev/releases/sam-witteveen-nvidia-dory-nov2/

## In short

Sam Witteveen explains NVIDIA's Dory, an open 4B looped transformer that reruns a shared block of layers 1, 2 or 4 times per token, and tests it against Gemma 4 26B.

## What it is

'Dory - NVIDIA's Looped Transformer' went up on Sam Witteveen's channel on 2 November 2026. The video covers Nemotron Labs Dory 1.0 4B, a research release NVIDIA put on Hugging Face on 16 October 2026, with a live demo of the model at 1, 2 and 4 loops.

## How it works

Dory splits its layers into an 8-layer input block, a 20-layer shared recurrent block and an 8-layer output block, and runs the shared block several times per token. The video explains this loop design, compares a per-loop KV cache with a last-loop cache that keeps memory flat, and shows how to deploy the model.

## Why it matters

A looped model lets one checkpoint trade speed for accuracy at serving time, with no second model to download. The video shows where that pays off and where Dory still loses to larger models.

## Who it's for

developers running small local models

## Links

- [Watch on YouTube](https://www.youtube.com/watch?v=vSCYL8ySh9g)
- [Dory on Hugging Face](https://huggingface.co/collections/nvidia/nemotron-labs-dory)
- [Dory Model Card](https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16)
- [Sam Witteveen](https://www.youtube.com/channel/UC55ODQSvARtgSyc8ThfiepQ)