Nemotron Labs Dory 1.0 4B
Small open looped model that reuses a shared block to think harder.
Dory 1.0 4B is an open 4B-parameter model from NVIDIA’s Nemotron Labs that runs a shared middle block of layers several times per token (one to four loops; NVIDIA recommends two), so one set of weights trades speed for accuracy. It is trained on English and 43 programming languages, always reasons, and supports tool calling.
- Provider
- NVIDIA
- Type
- Language model
- Released
- Oct 16, 2026
- Context window
- 262K tokens
- Input
- text
- Output
- text
- Open weights
- Yes
- License
- OpenMDW 1.1
Best for
- Reasoning
- Coding
- Local & on-device
Strengths
- Looped (recurrent-depth) design
- 256K-token context
- Open weights
More from NVIDIA
- Nemotron 3 Super — NVIDIA. Open hybrid Mamba-Transformer MoE for multi-agent applications.
- Nemotron 3.5 Lightning — NVIDIA. Small open 30B-A3B model tuned for low latency.
- Nemotron 3.5 Super VL — NVIDIA. Open multimodal MoE for documents, GUIs and video.
- Nemotron 3.5 Bolt — NVIDIA. Compact open vision-language MoE for running agents locally.
- Nemotron 3 Ultra — NVIDIA. NVIDIA’s largest open model: hybrid Mamba-Transformer MoE, 55B active.
- Nemotron 3 Embed 1B — NVIDIA. Small open embedding model for high-throughput enterprise retrieval.
- Cosmos 3 — NVIDIA. NVIDIA’s open omni-modal world foundation model for physical AI.
- Isaac GR00T N1.7 — NVIDIA. Open 3B cross-embodiment VLA for humanoids and manipulators.