Dextra
Cross-embodiment VLA manipulation foundation model + fine-tuning toolkit.
Products
The GripSim manipulation stack
Five products, one idea: give any robot hand human-level dexterity. Pick a layer or use them together.
Powering robots on the NVIDIA stack
NVIDIA InceptionIsaac LabGR00TJetson ThorOmniverseCosmos
All productsModelsRuntimeDataAgentic
Tactyl
Tactile-and-vision contact world model that predicts grasp outcomes.
Reflex
On-robot edge runtime for low-latency closed-loop manipulation.
Simdex
Domain-randomized manipulation simulation + data-generation engine.
Sequa
Agentic skill API that sequences and recovers grasps into tasks.
Why a stack
Each layer makes the others stronger
Models, world model, runtime, data engine, and agent — designed to compound.
Shared policy
Dextra is the brain every product builds on.
Closed loop
Tactyl + Reflex turn predictions into real-time control.
Compounding data
Simdex + deployments feed one growing flywheel.
Adoption
Start with one, grow into the stack
1
Pilot with Dextra
Prove dexterity on your hardware with the foundation model.
2
Add Reflex
Ship it on-robot at control-loop latency.
3
Layer Tactyl & Sequa
Add touch prediction and multi-step task recovery.
4
Scale with Simdex
Generate twin-based fine-tuning data as you expand.
Same space
Five angles on one manipulation brain
Every product stays inside the manipulation layer — variations and components of one idea, never a different business.
- Dextra — foundation policy
- Tactyl — contact world model
- Reflex — edge runtime
- Simdex — data engine
- Sequa — agentic skills
5
Products, one brain
1
Shared data flywheel
∞
Hands supported over time
Builders
Teams compose the stack their way
“We started with Dextra and Reflex, then added Sequa for kitting.”
“Simdex let us train before our hardware even shipped.”
“Tactyl cut our slip rate on deformables dramatically.”
Products FAQ
Common questions
Can I buy just one product?
Yes — each is usable on its own, and they integrate cleanly when you want more.
Are products priced separately?
Runtime is per-robot/fleet; the skill API is usage-based; fine-tuning and pilots are project-based.
Do they all need NVIDIA hardware?
Yes — training uses DGX/HGX and edge inference uses Jetson; GPUs are essential to each.
Why GripSim
Built for production robotics
The manipulation layer teams ship with, not a research demo.
Cross-embodiment by design
One policy runs across many grippers and hands — no per-robot rewrite.
Closed-loop with touch
Tactyl predicts slip and force so grasps react before they fail.
Runs on the robot
Reflex executes on Jetson at control-loop latency, with safety limits.
A moat that compounds
Every deployment feeds the shared model through the data flywheel.
Stay in the loop
Manipulation research, in your inbox
Benchmarks, sim-to-real notes, and product updates. No hype, no spam.
Find your manipulation layer
Tell us your robot and task — we'll map the right products.