NVIDIA's Dory1.0-4B Looped Model Lets You Pick Inference Depth at Runtime
lmoroney · x · 2026-10-20
NVIDIA Nemotron Labs released Dory1.0-4B, a looped model that reuses a shared block of 20 layers and lets you choose 1-4 passes at run time. A single checkpoint offers a fast setting for simple calls and a deeper one for hard coding tasks — no retraining, no model swap.
Laurence Moroney walks through how looping works, what NVIDIA reports, and what he'd want to measure before trusting it in an agent.
Related event: NVIDIA Open-Sources Dory1.0-4B Looped Model with Adjustable Inference Depth(2 posts)→
More from Models
- New decision model basal-omni-1.0-ultra claims near-frontier intelligence at much lower cost — tristanbob · 2026-10-20
- Reasoning models can steer their chain of thought far better than earlier tests suggested, paper finds — dair_ai · 2026-10-20
- Same sources, same question: how Claude, NotebookLM and ChatGPT differ on one knowledge base — kbavandi · 2026-10-20
- 434 Days Later: ChatGPT and Claude Still Can't Render JavaScript Content — gaganghotra_ · 2026-10-20
- Dream-1 claims new Pareto frontier for robot control: 16x cheaper, 1.7x faster, 59 vs 48 on RoboDojo — scott_e_reed · 2026-10-20
- $0.08 test finds DeepSeek's new search API scores by host, leaks spam despite filters — gaganghotra_ · 2026-10-20