Gestalt Robotics Reveals the Forma H7: A Home Robot That Actually Lears Your Household

Why Every Home Robot Until Now Has Been a Disappointment

Walk into any electronics retailer in 2027 and you will find a dozen robots that promise to help around the house. Nearly all of them operate on a simple loop: detect an object, execute a pre-programmed action, repeat. They cannot generalize. They cannot learn from context. They cannot handle a slightly crumpled shirt, a wine glass in an unusual spot, or a dining table with mismatched chairs. The result is a category full of expensive devices that can do exactly one thing, in exactly one way, until they get stuck.

Gestalt Robotics, a two-year-old spin-out from Carnegie Mellon's Robotics Institute, believes it has built something categorically different. Today the company unveiled the Forma H7, a home robot that it describes as "the first household robot that genuinely understands your home."

The Core Innovation: Hierarchical Sensory Grounding

The key technical breakthrough in the Forma H7 is a system Gestalt calls Hierarchical Sensory Grounding (HSG). Rather than programming the robot with fixed task sequences, HSG gives the robot a layered model of what a home is—and lets it figure out how to accomplish goals within that model.

At the base layer, the robot maintains a real-time 3D semantic map of every room it has explored: where the couch is, where the silverware drawer is, which items are fragile, which surfaces are heat-sensitive. Above that, HSG maintains a model of household activity sequences—what a dinner-setting looks like, what laundry folding entails, what a "tidy living room" means in the context of this specific family's habits. The robot learns this model through passive observation over the first two weeks of deployment, and continuously refines it.

The result is a robot that, when asked to "set the table for four," will: infer which table the household uses for meals (not the coffee table), retrieve the correct number of plates from the correct cabinet (learned from watching you do it), and place napkins and utensils in the positions consistent with what it has observed—without any explicit programming for your particular home.

Hardware Platform

The Forma H7 stands 1.2 meters tall and moves on a quad-track drivetrain capable of traversing thresholds up to 3cm and carpet-pile up to 2.5cm. Its manipulation system uses a pair of 7-degree-of-freedom arms with custom compliant grippers capable of handling objects from a wine glass to a cast-iron skillet. The robot's sensing suite includes:

• 4 × RGB-D cameras (front, rear, wrist × 2)
• A 128-beam LIDAR array in the head unit
• Capacitive touch sensors on all gripper fingers
• A microphone array for spatial voice localization

The compute unit runs a custom 48-TOPS neural accelerator on a 4nm process, with all household model inference performed on-device. No cloud connectivity is required for core functionality, though optional cloud sync allows users to migrate their household model to a new unit.

Pricing and Availability

The Forma H7 ships in three configurations:

• Forma H7 Base: $12,400 — includes core HSG, laundry folding, table setting, and dish-loading skills
• Forma H7 Plus: $16,800 — adds cooking assistance (ingredient prep, pot monitoring, timer management)
• Forma H7 Elite: $22,000 — includes the full skill library, dual-recipe learning from observation, and priority software updates

All units begin shipping in March 2028 to customers in the United States, Canada, the United Kingdom, Germany, France, and Japan. The company says international expansion is planned for late 2028.

The Test That Matters

The true measure of the Forma H7 will be longitudinal: can it handle the chaos of a real home over months and years? Gestalt is betting that its learning-based approach—rather than scripted task execution—will allow the robot to adapt to the disorder and variation that breaks every competitor. The first independent long-term stress tests begin in January 2028. We will be watching closely.