AnswersPublished16 sources
How Bear Robotics’ KINO-2 Turns Human Demonstrations Into Robot Skills
KINO 2 turns workers’ task demonstrations into manipulation skills that Bear trains, tests on its KR1 humanoid, and takes into customer pilots.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- KINO 2 turns workers’ task demonstrations into manipulation skills that Bear trains, tests on its KR1 humanoid, and takes into customer pilots.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How does Bear Robotics’ KINO-2 robot-learning platform, launched on October 21, 2026, at NVIDIA GTC Berlin, turn human task demonstrations i. Article summary: KINO-2 is Bear Robotics’ learning-and-deployment platform for turning human task demonstrations into manipulation skills on its KR1 humanoid, announced on October 21, 2026, at NVIDIA GTC Berlin. Its central aim is to con. Topic tags: general, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, clic
Bear Robotics’ KINO-2 is a learning-and-deployment platform designed to teach its KR1 humanoid to handle objects, extending the company’s automation beyond moving goods. Announced at NVIDIA GTC Berlin on October 21, 2026, the platform uses human task demonstrations to train and refine models, tests the resulting skills on KR1, and advances them into customer pilots. 4
20
From a human demonstration to a robot pilot
Bear describes a process that starts by collecting demonstrations of a task. The company uses those examples to train and refine models, evaluates the resulting skills on KR1, and then moves them into a customer pilot. The goal is to translate demonstrated work into deployable object-handling capabilities that can connect with existing robot mobility and workflow systems. 10
20
This builds on Bear’s earlier KINO learning program. In 2026, the company acquired Kinisi Robotics, bringing the KR1 humanoid and the original KINO program into its portfolio; KINO-2 extends that work toward commercial deployment. 10
21 Bear says it has deployed more than 16,000 robots, mostly service robots, at over 5,000 customer locations.
20
What NVIDIA’s software contributes
KINO-2 uses NVIDIA Isaac OS as the perception runtime for KR1. Bear says Isaac OS, which incorporates Holoscan SDK 5.0, connects perception processing and model inference on NVIDIA Jetson AGX Thor. The company also reports using TensorRT and model quantization to optimize inference. 4
Bear reports that Isaac GR00T inference time fell from about 500 milliseconds to about 60 milliseconds. For its own model, the reported time dropped from 72.3 to 17.2 milliseconds. These are company-reported model-inference measurements—not end-to-end task times. 4
6
NVIDIA describes Bear using KINO-2 to post-train GR00T for warehouse, returns-processing, and industrial workflows. 9 The available reporting does not detail how KINO-2 supports external foundation models or quantify how much task-specific engineering it removes, so those capabilities should not be assumed from the platform’s general description alone.
What the early applications and results show
Bear describes warehouse packing as an example of manipulation connected to mobility: KR1 can learn to pick bottles and fill a crate, while Bear’s broader systems support moving goods through a workflow. Customer pilots are underway, but the published material cited here does not provide full-cycle task times, manipulation accuracy, or long-term reliability results. 10
20
That distinction matters when interpreting the speed figures. Faster model inference may help a robot respond, but it does not by itself establish that a complete warehouse or returns task is faster, more accurate, or reliable over time. KINO-2’s stated advance is the demonstration-to-pilot pipeline; evidence of production performance will depend on results from real-world tasks.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "How Bear Robotics’ KINO-2 Turns Human Demonstrations Into Robot Skills"?
KINO 2 turns workers’ task demonstrations into manipulation skills that Bear trains, tests on its KR1 humanoid, and takes into customer pilots.