Sidekick Robotics annual ranking

2026

The Annual State of Spatial Intelligence Ranking

Sidekick Robotics' inaugural list of the 10 AI models defining spatial intelligence in 2026.

Spatial intelligence / launch film2026

In 2026, spatial intelligence became a defining frontier for AI: systems that can perceive, reason about, and act in the three-dimensional world.

Sidekick Robotics reviewed more than 700 models to identify the 10 shaping the field, from creative tools to the future of physical work.

The 2026 list

The models shaping spatial intelligence

World Labs' Atlas tops the inaugural list, followed by Skild AI's S1, Generalist AI's GEN-1.5 and Physical Intelligence's π0.7.

  1. 01
    Atlas

    World Labs

    World Labs

    An omni world model that generates up to one minute of 1440p video with user-directed camera paths from a reference image, while maintaining 3D scene consistency. Introduced in September 2026.

  2. 02
    S1

    Skild AI

    Skild AI

    A robotic foundation model for in-context learning: a single video demonstration can prompt long-horizon manipulation tasks, reportedly up to 10 minutes, without fine-tuning or weight updates.

  3. 03
    GEN-1.5

    Generalist AI

    Generalist AI

    An embodied foundation model that uses short sensorimotor demonstrations as physical prompts, enabling one-shot or few-shot learning for new robot tasks without gradient updates.

  4. 04
    π0.7

    Physical Intelligence

    Physical Intelligence

    A robotic foundation model focused on compositional generalization: it recombines learned skills for novel multi-stage tasks, including appliance interaction and zero-shot cross-embodiment laundry demonstrations.

  5. 05
    Marble

    World Labs

    World Labs

    A generative world model that creates persistent, explorable 3D environments from text, images, video or 3D layouts; launched generally in November 2025 with free and paid plans.

  6. 06
    Gemini Robotics ER 2

    Google DeepMind

    Google DeepMind

    A high-level vision-language model for robot spatial reasoning and task orchestration. It can break complex workflows into subtasks and delegate actions to lower-level robot models through the Gemini API.

  7. 07
    Cosmos 3

    NVIDIA

    NVIDIA

    An open family of omnimodal world models for physical AI, combining world understanding, simulation and generation across modalities including text, images, video, sound and actions.

  8. 08
    GPT-6 Astra

    OpenAI

    OpenAI

    A general-purpose model evaluated on robot-control tasks, where reported RoboDojo results showed strong semantic and open-task performance alongside weaker precision control. Released in September 2026.

  9. 09
    Genie 3

    Google DeepMind

    Google DeepMind

    A world model for generating interactive environments that users can explore and remix from text or image prompts. Project Genie made the experience available through Google Labs in January 2026 to eligible U.S. subscribers.

  10. 10
    Isaac GR00T N1.7

    NVIDIA

    NVIDIA

    An open foundation model for humanoid robot reasoning and skills, built for robot action generation and positioned within NVIDIA's broader physical-AI model ecosystem.

How we ranked / 2026

An editorial view informed by market heat.

Our team reviewed more than 700 robot models, world models and related research from about 190 labs. Each leading model received a score from 1 to 5 across four signals.

  • 012026 launchHow new and significant the release is.
  • 02Market attentionPress, social and market reaction.
  • 03Access and adoptionWho can use it today through open weights, an API or deployments.
  • 04Maker momentumFunding, valuation and deals over the past 12 months.

The final order reflects our editors' judgment, informed by those scores. The #1 spot is our editors' pick: Atlas, the newest model from World Labs, the company most closely tied to spatial intelligence. Data is current as of October 6, 2026.

About Sidekick Robotics

The infrastructure behind spatial intelligence.

Sidekick Robotics builds AI infrastructure for embodied AI, enabling enterprises to develop, deploy and own robotic intelligence for the physical world.

We publish the State of Spatial Intelligence Ranking every year. Building a model that belongs on next year's list? Tell us about it.