DEPLOYDatabase

Brain

DYNA-2

DYNA-2 is Dyna Robotics' flagship World-Action Model (WAM), pre-trained on over 1 million hours of human egocentric video — roughly 170 years of continuous waking experience. Unlike VLA (Vision-Language-Action) models, DYNA-2 is built on a video-diffusion backbone that jointly denoises future video and future actions, enabling physical intuition and spatial reasoning to transfer directly from human video to robot hardware without any robot data in pre-training.

Announced August 10, 2026, it demonstrates the first human-to-robot scaling law: robot performance improves smoothly and predictably with every added hour of human data, without hitting plateaus. DYNA-2 enables zero-shot performance at production-level speeds across new deployment sites, and can be adapted to new robot platforms with just a few hours of local fine-tuning. In one benchmark, 13 minutes of data was sufficient for DYNA-2 to command five-fingered robot hands to twist open a bottle cap.

World model · Closed · Powers 1 robot

Machine-readable surfaces

Key facts

Brain type
World-Action Model (WAM), video-diffusion backbone
Pre-training data
1,000,000+ hours of human egocentric video (~170 years of continuous waking experience)
Architecture
Dual next-frame and next-action world-modeling; jointly denoises future video and future actions
Paradigm shift
First human-to-robot scaling law, performance scales smoothly with human video data, no plateaus
Fine-tuning
Adapts to new robot platforms with hours of local data (13 min for bottle-cap task)
Key result
133% improvement on instruction-following tasks via video co-training; zero-shot production-level performance
Announced
August 10, 2026
Predecessor
DYNA-1 (foundation-model, April 2025)
World-action model
Dyna Robotics presents DYNA-2 as a World-Action Model trained on more than one million hours of human video (Dyna-2).
Robot adaptation claim
Dyna Robotics says DYNA-2 is designed to transfer from human video data to robot action and adapt across robot platforms (Dyna-2).

Developed by (1)

Powers (1)

Where DYNA-2 runs

View the global map

Deployments of robots powered by DYNA-2, by place and type.

Free to use on your site, with a link back to DEPLOY.

Mentioned on (4)

Registry pages whose text names DYNA-2.

Companies (1)

Dyna Robotics

Robots and products (2)

DYNA 2.1, Sophy Swiftfold

AI systems (1)

DYNA-1

Common questions

What is DYNA-2?
DYNA-2 is Dyna Robotics' flagship World-Action Model (WAM), pre-trained on over 1 million hours of human egocentric video — roughly 170 years of continuous waking experience. Unlike VLA (Vision-Language-Action) models, DYNA-2 is built on a video-diffusion backbone that jointly denoises future video and future actions, enabling physical intuition and spatial reasoning to transfer directly from human video to robot hardware without any robot data in pre-training. Announced August 10, 2026, it demonstrates the first human-to-robot scaling law: robot performance improves smoothly and predictably with every added hour of human data, without hitting plateaus. DYNA-2 enables zero-shot performance at production-level speeds across new deployment sites, and can be adapted to new robot platforms with just a few hours of local fine-tuning. In one benchmark, 13 minutes of data was sufficient for DYNA-2 to command five-fingered robot hands to twist open a bottle cap.
Which robots run on DYNA-2?
1 robot model on the DEPLOY registry runs on DYNA-2: Sophy Swiftfold.
Who developed DYNA-2?
DYNA-2 is credited to Dyna Robotics on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is DYNA-2 open source?
No. DYNA-2 is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.
What type of AI is DYNA-2?
DYNA-2 is a world model on the DEPLOY registry.

Sources (3)

  1. https://www.dyna.co/dyna-2
  2. https://www.prnewswire.com/news-releases/dyna-robotics-unveils-dyna-2-world-action-model-demonstrating-first-true-scaling-law-in-robotics-powered-entirely-by-human-data-302847114.html
  3. https://www.dyna.co/
Methodology: Verified · 3 sources (1 primary) · last reviewed 2026-10-09

Verification posture

Verified

High confidence

Review state

Stable

Last reviewed 2026-10-09

Sources by quality tier

2
unclassified
Unclassified source
1
primary-company-ir
Company IR disclosure

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for DYNA-2.

Canonical ID f8b378da-d4b0-4c84-af50-d7ce909ba834