--- title: "Gemini Robotics 2 Whole-Body Control Guide (2026)" canonical: https://wavect.io/blog/gemini-robotics-2-whole-body-control/ language: en description: "Gemini Robotics 2 explained for product teams: three models, whole-body benchmarks, access, safety limits and a practical robotics pilot plan." image: "https://wavect.io/img/blog/headers/header_gemini-robotics-2-whole-body-control.png" --- [**Back**](/blog/overview/) [![Kevin Riedl](/img/team/kevin.webp)](/team/kevin-riedl/) [Kevin Riedl](/team/kevin-riedl/) https://linkedin.com/in/wsdt 11 min read · 7 Aug 2026 [**Next**](/blog/ai-agent-pilot-30-60-90-days/) # Gemini Robotics 2: What Whole-Body Control Changes for Product Teams TL;DR Gemini Robotics 2 is a three-model physical AI stack: a VLA for direct whole-body action, Gemini Robotics ER 2 for planning and progress tracking, and an on-device VLA for low-latency local control. The important shift is coordinated movement from feet to fingertips, not a single dexterity demo. Google DeepMind reports 92% success at unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks, so the release is a capability step rather than proof of general production reliability. Gemini Robotics ER 2 is available through the Gemini API and AI Studio, while the direct-control VLA models remain limited to early-access or trusted-test partners. Teams can pilot embodied reasoning, tool orchestration, video-based progress checks and simulation now. A physical deployment still needs deterministic low-level safety, hardware interlocks, human stop authority and task-specific evaluation. **Gemini Robotics 2 moves Google DeepMind's robotics stack from upper-body manipulation to coordinated control from feet to fingertips.** The commercial implication is larger than a humanoid learning one more trick. A robot can now combine locomotion, balance, reaching and manipulation while a separate reasoning model plans a multi-step job and checks whether each step worked. The release does not make humanoid deployment plug and play. Google reports a 92% success rate for unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks. Direct-control models also remain limited to selected partners. The practical question is therefore not “can we buy a general-purpose robot worker?” It is **which part of this stack can we test now, against which workflow, with which safety boundary?** This guide answers that decision. It separates the three models, reads the benchmark beyond its best number, explains current access and gives product teams a bounded pilot plan. For the broader rollout discipline, pair it with our [30/60/90-day AI agent pilot plan](/blog/ai-agent-pilot-30-60-90-days/). ## What is Gemini Robotics 2? **Gemini Robotics 2 is a family of three connected physical AI models, not one monolithic robot brain.** Google DeepMind's [official Gemini Robotics 2 release](https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/) divides perception, planning and action across models with different deployment characteristics. | Model | Role in the stack | Current access | | --- | --- | --- | | Gemini Robotics 2 | Vision-language-action model that converts images and instructions into motor actions for full humanoids and bi-arm robots | Early-access partners | | Gemini Robotics ER 2 | Embodied reasoning VLM that understands video, plans multi-step tasks, calls tools and tracks completion | Gemini API, Google AI Studio and private enterprise preview | | Gemini Robotics On-Device 2 | Efficient VLA for local robot inference where network latency or connectivity is unacceptable | Trusted testers | This separation matters. The reasoning model can run as the high-level orchestrator while a VLA, navigation API or conventional controller executes a bounded action. Teams can improve or replace one layer without pretending one foundation model should own every joint, safety stop and business decision. ## Why whole-body control is the real jump The previous generation mainly controlled a humanoid's upper body for tabletop tasks. Gemini Robotics 2 expands the learned action policy to the entire humanoid. In DeepMind's Apollo 2 demo, one instruction causes the robot to walk to a table, pick up a watering can, step toward shelving and place it in a lower bin. Every stage changes reach, balance and the next useful camera view. That coordination is the bottleneck in environments built for people. A fixed arm can be highly capable inside a designed cell. A humanoid in a stockroom, lab or maintenance area must move between work surfaces, adjust its stance, avoid people and recover when an object is not where the plan expected it. Whole-body control reduces the handoffs between separately programmed locomotion and manipulation routines. It does not remove the rest of the robotics stack. Hardware-specific control loops still own collision-free motion, balance, force limits and emergency behavior. The foundation model expands which goals the system can interpret and which movements it can compose. It should not become the only safety layer. ## What does the 92% light-bulb result actually prove? **It proves meaningful task-specific dexterity, not general hand reliability.** The launch evaluation reports individual success rates for five multi-finger tasks on Apollo 2 with a 22-degree-of-freedom hand. | Task | Reported success | What a buyer should infer | | --- | --- | --- | | Unscrew a light bulb | 92% | Strong performance on the best demonstrated multi-finger task | | Screw in a light bulb | 36% | The reverse action is not equally solved | | Tie a trash bag | 44% | Deformable objects remain difficult | | Close a ziplock bag | 40% | Fine alignment and force control still fail often | | Use a dustpan | 32% | Coordinated tool use is not yet reliable enough for unattended work | Other release charts show 45.7% for picking an object from the floor with one Apollo configuration, compared with 68.4% from a table and 76.3% from a shelf. That is exactly why a pilot needs a representative task distribution rather than a highlight reel. Measure the awkward angles, dropped objects, clutter, lighting changes, people entering the workspace and recovery after partial failure. ## How do the three models work together? Gemini Robotics ER 2 acts as the high-level agent. It consumes continuous video and instructions, determines the steps, calls lower-level tools, observes progress and either continues, retries or asks for help. Google's [developer release for Gemini Robotics ER 2](https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/) says developers can declare VLA models or navigation APIs as tools and stream video, audio or text through the Gemini Live API. A production-minded control path should look like this: 1. **Observe:** cameras, robot state and task context enter a time-stamped perception stream. 2. **Reason:** ER 2 selects the next bounded step, checks constraints and decides whether clarification is required. 3. **Execute:** a VLA or deterministic robot API performs the approved action. 4. **Guard:** low-level controllers, safety-rated sensors and hardware interlocks can stop motion independently. 5. **Verify:** video progress checks and deterministic task signals confirm the outcome. 6. **Escalate:** uncertainty, repeated failure or a person entering the zone returns control to a human. This is an agent architecture with physical consequences. Logging must preserve the instruction, observations, proposed tool call, constraint decision, executed motion, stop events and final verification. Without that trace, the team cannot explain a failure or decide whether a new model version improved the system. ## Can a company use Gemini Robotics 2 today? **Yes for embodied reasoning experiments; usually not yet for the direct whole-body model.** The [Gemini Robotics ER 2 model card](https://deepmind.google/models/model-cards/gemini-robotics-er-2/) lists the Gemini API and AI Studio as distribution channels. It also tells users to use discretion in production, commercial or public environments and not to use the model for safety-critical applications. | Your starting point | What to do now | What not to assume | | --- | --- | --- | | No robot hardware | Prototype video understanding, tool orchestration and progress checks in simulation or recorded workflows | Simulation success will transfer unchanged to hardware | | Existing robot with APIs | Place ER 2 above narrow, tested skills and keep human approval on motion | The reasoning model replaces certified safety logic | | Bi-arm research platform | Apply for trusted-tester access and prepare task demonstrations, telemetry and baselines | Fewer than 200 examples guarantees a target success rate | | Humanoid deployment plan | Run a workflow and safety discovery before hardware procurement | A polished Apollo demo is a general availability commitment | If you are choosing between custom integration, a packaged platform and waiting, our [custom software versus off-the-shelf decision guide](/software-development-guide/custom-software-vs-off-the-shelf/) provides the commercial framework. In robotics, add hardware availability, retraining data, safety ownership and vendor access to the usual cost and lock-in questions. ## Why the on-device transfer result may matter most commercially Cloud reasoning is useful when the network and response time fit the task. Motion control often cannot accept that dependency. Gemini Robotics On-Device 2 is designed to run locally and DeepMind says it can adapt to a new bi-arm embodiment with a few hours of data, typically fewer than 200 examples. The [On-Device 2 model card](https://deepmind.google/models/model-cards/gemini-robotics-on-device-2/) adds the boundary that headlines omit. The model is currently available only to trusted testers, is limited on out-of-distribution tasks and high-degree-of-freedom robots, and was evaluated mainly for standing bi-arm manipulation. Mobile and whole-body safety risks sit outside that model card's evaluation scope. So “under 200 examples” is a data-efficiency starting point, not a deployment quote. A buyer still needs to ask how examples are collected, who labels success, how many failure and edge cases are included, which sensors change, what compute runs on the robot and what happens when the new body produces a state outside the adaptation set. ## The safety result argues for layers, not autonomy DeepMind's [Gemini Robotics 2 safety report](https://storage.googleapis.com/deepmind-media/gemini-robotics/Gemini-Robotics-2-Safety.pdf) is useful precisely because it shows the trade-offs. In its human-proximity benchmark, no tested model achieved both near-zero missed hazards and near-zero unnecessary stops. Models tuned to interrupt less could miss more genuine hazards. The report recommends deterministic low-level safety guardrails alongside frontier-model perception. For a commercial pilot, define four independent controls before the first live movement: - **Physical boundary:** speed, force, workspace and payload limits enforced below the model. - **Semantic boundary:** forbidden tasks, tool allowlists and required clarification for ambiguous instructions. - **Human boundary:** visible stop authority and approval for new tasks, tools or operating zones. - **Evidence boundary:** task-level success, collision, intervention, false-stop and recovery metrics recorded per run. A model recognizing a nearby person is valuable. It is not a substitute for safety-rated sensing and machinery risk assessment. ## A 90-day Gemini Robotics 2 pilot plan 1. **Days 0 to 30, select the workflow.** Choose one repetitive task where variable scenes make fixed automation expensive. Record the human baseline, task frequency, intervention cost and unacceptable failure. Map every physical hazard and decide whether the pilot stays in simulation. 2. **Days 31 to 60, build the digital control boundary.** Integrate ER 2 with mocked or sandboxed robot tools. Stream representative video, require structured tool calls, add ambiguity tests and log progress. Run a fixed evaluation set before any live control. 3. **Days 61 to 90, supervise limited motion.** If the safety case permits, connect only reviewed skills at reduced speed inside a controlled zone. A human approves actions. Track completion, recovery, intervention, false stops, latency and cost per accepted task. The day-90 decision is expand, redesign or stop. Do not scale because the robot completed the best demo once. Scale when the measured task distribution clears a pre-agreed threshold and the team can explain and recover from every important failure mode. ## What should buyers ask a robotics AI partner? - Which model is actually available to us today: ER 2 through the API, or a VLA through a partner program? - Which benchmark task resembles our workflow, and what is the complete success distribution rather than the best result? - What remains deterministic below the model, including balance, collision avoidance, force and emergency stop? - How are adaptation examples collected, versioned and approved? - What data leaves the site, and which inference remains on-device? - How do we replay failures and compare model, prompt or controller versions? - Who owns the integration, training data, telemetry and rollback path if model access changes? Wavect's [AI product engineering service](/services/artificial-intelligence/) can help turn the reasoning, integration and evaluation layer into a bounded pilot your team owns. Our [Twinsoft AI case study](/case-studies/twinsoft-ai/) shows the product and engineering discipline behind moving an AI system from concept to reliable workflow. If you need an independent feasibility decision before hardware spend, [book a physical AI discovery call](/contact/). ## Frequently Asked Questions ### What is Gemini Robotics 2? Gemini Robotics 2 is Google DeepMind's three-model physical AI family: a whole-body vision-language-action model, the Gemini Robotics ER 2 embodied reasoning model and an efficient on-device VLA. Together they separate task planning, progress tracking and physical action. ### What is whole-body control in Gemini Robotics 2? Whole-body control means one learned action model coordinates a humanoid's locomotion, torso, arms and hands to complete a goal. The previous Gemini Robotics generation mainly controlled upper-body tabletop tasks. ### Is Gemini Robotics 2 available to developers? Gemini Robotics ER 2 is available through the Gemini API and Google AI Studio. The direct Gemini Robotics 2 VLA is offered to early-access partners, and On-Device 2 is limited to trusted testers at the time of writing. ### Did Gemini Robotics 2 achieve 92% success? It achieved a reported 92% on the specific task of unscrewing a light bulb. Other multi-finger results were lower, including 36% for screwing in a bulb, 44% for tying a trash bag, 40% for closing a ziplock bag and 32% for using a dustpan. ### Can Gemini Robotics 2 adapt to a new robot with fewer than 200 examples? DeepMind says On-Device 2 can adapt to new bi-arm embodiments in a few hours, typically using fewer than 200 examples. That is not a guarantee of production success and the model card notes limits on out-of-distribution tasks and high-degree-of-freedom robots. ### Is Gemini Robotics 2 safe for production robots? It should not be the only safety system. Google's model card excludes safety-critical use, and its safety report recommends deterministic low-level guardrails. Physical deployments still need safety-rated sensing, hardware limits, emergency stops, human authority and task-specific risk assessment. ### What is the best first commercial use of Gemini Robotics ER 2? Start with a bounded, reversible workflow such as video-based progress checking or orchestration of existing robot APIs in simulation. This tests reasoning and tool use without immediately granting a model uncontrolled physical authority. ## Primary sources and research boundary 1. Google DeepMind, Gemini Robotics 2 release; model family, access, whole-body demos, dexterity benchmarks and adaptation claims. 2. Google, Gemini Robotics ER 2 developer release; API access, tool orchestration, streaming and progress-evaluation details. 3. Google DeepMind, Gemini Robotics ER 2 model card; distribution, intended use and production limitations. 4. Google DeepMind, Gemini Robotics On-Device 2 model card; trusted-tester access, evaluation scope and known limitations. 5. Google DeepMind, Gemini Robotics 2 safety report; human-proximity, uncertainty and layered-safety findings. *Status checked 7 August 2026. This analysis uses public release material, model cards and safety evaluations. We did not receive model access, operate Apollo 2 or independently reproduce the robotics benchmarks. Reported success rates are vendor evaluations, not Wavect test results.* ## Final thoughts Gemini Robotics 2 matters because it moves the fast-changing part of robotics from isolated manipulation toward a layered control system that can reason, move and verify across a complete task. Whole-body coordination and rapid transfer could reduce the custom programming needed for every new robot and workflow. The release also shows why buying decisions need the full benchmark. A 92% result can coexist with 32% to 44% on nearby tasks, public access to the reasoning layer can coexist with restricted access to the action models, and better semantic safety can coexist with a continuing need for deterministic guards. Pilot the narrow workflow, measure the complete distribution and let evidence earn every increase in physical autonomy. ## You may also like.. [**AI Agent Pilot in 30/60/90 Days** Use a measured rollout plan for the reasoning and orchestration layer behind a physical AI pilot.](/blog/ai-agent-pilot-30-60-90-days/) [**AI Enablement vs an In-House AI Hire** Compare a focused external build-and-handover engagement with creating a permanent internal role.](/compare/ai-enablement-vs-in-house-ai-hire/) Models and infrastructure ## Continue through this cluster [Start with the cornerstone**Self-Hosting LLMs in the EU: When Open Weights Actually Pay Off**](/blog/self-hosting-llms-eu-cost/) - [pdf-inspector Review: Route PDFs Before OCR](/blog/pdf-inspector-ocr-routing/) - [Local Multimodal AI Coding Assistant: Voice, OCR and Privacy](/blog/local-multimodal-ai-coding-assistant/) - [DeepSeek V4 Flash 0731 on One AI PC: What Actually Works?](/blog/deepseek-v4-flash-0731-local-ai-pc/) - [Fine-Tune Gemma 4 Free with Unsloth and Colab](/blog/fine-tune-gemma-4-free-unsloth-colab/) - [Taalas HC1 Review: Is a Hardwired LLM ASIC Worth It?](/blog/taalas-hc1-llm-asic-review/) Inbox, without the noise ## Follow the work that matters to you Get a short email when we publish something new. Follow the whole blog or only the problems you care about. [**Back**](/blog/overview/) [![Kevin Riedl](/img/team/kevin.webp)](/team/kevin-riedl/) [Kevin Riedl](/team/kevin-riedl/) https://linkedin.com/in/wsdt 11 min read · 7 Aug 2026 [**Next**](/blog/ai-agent-pilot-30-60-90-days/) New posts by email × × Get new posts by email A short email when we publish. Free, no tracking. ## Structured Data ```json { "@context": "https://schema.org", "@graph": [ { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ], "employee": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "founder": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "legalRepresentative": [ { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } } ], "name": "Wavect GmbH", "subjectOf": { "@id": "https://wavect.io/verified-claims.json#dataset", "@type": "Dataset", "creator": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "description": "A machine-readable registry of quantitative and qualitative claims published by Wavect, with review dates, localized page appearances and public third-party citations where available.", "inLanguage": "en", "isAccessibleForFree": true, "license": "https://creativecommons.org/licenses/by/4.0/", "name": "Wavect verified publication claims", "url": "https://wavect.io/verified-claims.json" }, "url": "https://wavect.io/" }, { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Kevin Riedl", "sameAs": [ "https://www.wikidata.org/wiki/Q139796365", "https://www.linkedin.com/in/wsdt", "https://github.com/wsdt" ], "url": "https://wavect.io/team/kevin-riedl/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/team/christof-jori/#person", "@type": "Person", "jobTitle": "Managing Director", "name": "Christof Jori", "sameAs": [ "https://www.wikidata.org/wiki/Q139796367", "https://www.linkedin.com/in/jocr77/", "https://github.com/jo-chris" ], "url": "https://wavect.io/team/christof-jori/", "worksFor": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] } }, { "@id": "https://wavect.io/#website", "@type": "WebSite", "inLanguage": [ "en", "de", "es", "zh" ], "name": "Wavect", "potentialAction": { "@type": "SearchAction", "query-input": "required name=search_term_string", "target": { "@type": "EntryPoint", "urlTemplate": "https://wavect.io/search/?q={search_term_string}" } }, "publisher": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "url": "https://wavect.io/" }, { "@id": "https://wavect.io/blog/gemini-robotics-2-whole-body-control/#webpage", "@type": "WebPage", "dateModified": "2026-08-07", "inLanguage": "en", "isPartOf": { "@id": "https://wavect.io/#website", "@type": "WebSite" }, "lastReviewed": "2026-08-07", "url": "https://wavect.io/blog/gemini-robotics-2-whole-body-control/" } ] } ``` ```json { "@context": "https://schema.org", "@type": "BlogPosting", "abstract": "Gemini Robotics 2 is a three-model physical AI stack: a VLA for direct whole-body action, Gemini Robotics ER 2 for planning and progress tracking, and an on-device VLA for low-latency local control. The important shift is coordinated movement from feet to fingertips, not a single dexterity demo. Google DeepMind reports 92% success at unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks, so the release is a capability step rather than proof of general production reliability. Gemini Robotics ER 2 is available through the Gemini API and AI Studio, while the direct-control VLA models remain limited to early-access or trusted-test partners. Teams can pilot embodied reasoning, tool orchestration, video-based progress checks and simulation now. A physical deployment still needs deterministic low-level safety, hardware interlocks, human stop authority and task-specific evaluation.", "articleBody": " Blog overview/AI and agents/Models and infrastructure Gemini Robotics 2: What Whole-Body Control Changes for Product Teams TL;DR Gemini Robotics 2 is a three-model physical AI stack: a VLA for direct whole-body action, Gemini Robotics ER 2 for planning and progress tracking, and an on-device VLA for low-latency local control. The important shift is coordinated movement from feet to fingertips, not a single dexterity demo. Google DeepMind reports 92% success at unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks, so the release is a capability step rather than proof of general production reliability. Gemini Robotics ER 2 is available through the Gemini API and AI Studio, while the direct-control VLA models remain limited to early-access or trusted-test partners. Teams can pilot embodied reasoning, tool orchestration, video-based progress checks and simulation now. A physical deployment still needs deterministic low-level safety, hardware interlocks, human stop authority and task-specific evaluation. Gemini Robotics 2 moves Google DeepMind's robotics stack from upper-body manipulation to coordinated control from feet to fingertips. The commercial implication is larger than a humanoid learning one more trick. A robot can now combine locomotion, balance, reaching and manipulation while a separate reasoning model plans a multi-step job and checks whether each step worked. The release does not make humanoid deployment plug and play. Google reports a 92% success rate for unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks. Direct-control models also remain limited to selected partners. The practical question is therefore not “can we buy a general-purpose robot worker?” It is which part of this stack can we test now, against which workflow, with which safety boundary? This guide answers that decision. It separates the three models, reads the benchmark beyond its best number, explains current access and gives product teams a bounded pilot plan. For the broader rollout discipline, pair it with our 30/60/90-day AI agent pilot plan. What is Gemini Robotics 2? Gemini Robotics 2 is a family of three connected physical AI models, not one monolithic robot brain. Google DeepMind's official Gemini Robotics 2 release divides perception, planning and action across models with different deployment characteristics. ModelRole in the stackCurrent access Gemini Robotics 2Vision-language-action model that converts images and instructions into motor actions for full humanoids and bi-arm robotsEarly-access partners Gemini Robotics ER 2Embodied reasoning VLM that understands video, plans multi-step tasks, calls tools and tracks completionGemini API, Google AI Studio and private enterprise preview Gemini Robotics On-Device 2Efficient VLA for local robot inference where network latency or connectivity is unacceptableTrusted testers This separation matters. The reasoning model can run as the high-level orchestrator while a VLA, navigation API or conventional controller executes a bounded action. Teams can improve or replace one layer without pretending one foundation model should own every joint, safety stop and business decision. Why whole-body control is the real jump The previous generation mainly controlled a humanoid's upper body for tabletop tasks. Gemini Robotics 2 expands the learned action policy to the entire humanoid. In DeepMind's Apollo 2 demo, one instruction causes the robot to walk to a table, pick up a watering can, step toward shelving and place it in a lower bin. Every stage changes reach, balance and the next useful camera view. That coordination is the bottleneck in environments built for people. A fixed arm can be highly capable inside a designed cell. A humanoid in a stockroom, lab or maintenance area must move between work surfaces, adjust its stance, avoid people and recover when an object is not where the plan expected it. Whole-body control reduces the handoffs between separately programmed locomotion and manipulation routines. It does not remove the rest of the robotics stack. Hardware-specific control loops still own collision-free motion, balance, force limits and emergency behavior. The foundation model expands which goals the system can interpret and which movements it can compose. It should not become the only safety layer. What does the 92% light-bulb result actually prove? It proves meaningful task-specific dexterity, not general hand reliability. The launch evaluation reports individual success rates for five multi-finger tasks on Apollo 2 with a 22-degree-of-freedom hand. TaskReported successWhat a buyer should infer Unscrew a light bulb92%Strong performance on the best demonstrated multi-finger task Screw in a light bulb36%The reverse action is not equally solved Tie a trash bag44%Deformable objects remain difficult Close a ziplock bag40%Fine alignment and force control still fail often", "articleSection": "Engineering", "author": { "@id": "https://wavect.io/team/kevin-riedl/#person", "@type": "Person", "name": "Kevin Riedl", "sameAs": [ "https://www.wikidata.org/wiki/Q139796365", "https://www.linkedin.com/in/wsdt", "https://github.com/wsdt" ], "url": "https://wavect.io/team/kevin-riedl/" }, "citation": [ { "@type": "WebPage", "name": "official Gemini Robotics 2 release", "url": "https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/" }, { "@type": "WebPage", "name": "developer release for Gemini Robotics ER 2", "url": "https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/" }, { "@type": "WebPage", "name": "Gemini Robotics ER 2 model card", "url": "https://deepmind.google/models/model-cards/gemini-robotics-er-2/" }, { "@type": "WebPage", "name": "On-Device 2 model card", "url": "https://deepmind.google/models/model-cards/gemini-robotics-on-device-2/" }, { "@type": "WebPage", "name": "Gemini Robotics 2 safety report", "url": "https://storage.googleapis.com/deepmind-media/gemini-robotics/Gemini-Robotics-2-Safety.pdf" } ], "dateModified": "2026-08-07", "datePublished": "2026-08-07", "description": "Gemini Robotics 2 is a three-model physical AI stack: a VLA for direct whole-body action, Gemini Robotics ER 2 for planning and progress tracking, and an on-device VLA for low-latency local control. The important shift is coordinated movement from feet to fingertips, not a single dexterity demo. Google DeepMind reports 92% success at unscrewing a light bulb, but only 36% for screwing one in and 32% to 44% on several other multi-finger tasks, so the release is a capability step rather than proof of general production reliability. Gemini Robotics ER 2 is available through the Gemini API and AI Studio, while the direct-control VLA models remain limited to early-access or trusted-test partners. Teams can pilot embodied reasoning, tool orchestration, video-based progress checks and simulation now. A physical deployment still needs deterministic low-level safety, hardware interlocks, human stop authority and task-specific evaluation.", "headline": "Gemini Robotics 2: Whole-Body Control and the Pilot Decision", "image": "https://wavect.io/img/blog/headers/header_gemini-robotics-2-whole-body-control.svg", "inLanguage": "en", "keywords": "Gemini Robotics 2, Physical AI, Humanoid Robotics", "mainEntityOfPage": { "@id": "https://wavect.io/blog/gemini-robotics-2-whole-body-control/", "@type": "WebPage" }, "publisher": { "@id": "https://wavect.io/#organization", "@type": [ "Organization", "ProfessionalService", "LocalBusiness" ] }, "url": "https://wavect.io/blog/gemini-robotics-2-whole-body-control/", "wordCount": 2659 } ``` ```json { "@context": "https://schema.org", "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "item": "https://wavect.io/", "name": "Home", "position": 1 }, { "@type": "ListItem", "item": "https://wavect.io/blog/overview/", "name": "Blog overview", "position": 2 }, { "@type": "ListItem", "item": "https://wavect.io/blog/topics/ai-agents/", "name": "AI and agents", "position": 3 }, { "@type": "ListItem", "item": "https://wavect.io/blog/clusters/models-infrastructure/", "name": "Models and infrastructure", "position": 4 }, { "@type": "ListItem", "item": "https://wavect.io/blog/gemini-robotics-2-whole-body-control/", "name": "Gemini Robotics 2 Whole-Body Control Guide (2026) | ", "position": 5 } ] } ``` ```json { "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Gemini Robotics 2 is Google DeepMind's three-model physical AI family: a whole-body vision-language-action model, the Gemini Robotics ER 2 embodied reasoning model and an efficient on-device VLA. Together they separate task planning, progress tracking and physical action." }, "name": "What is Gemini Robotics 2?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Whole-body control means one learned action model coordinates a humanoid's locomotion, torso, arms and hands to complete a goal. The previous Gemini Robotics generation mainly controlled upper-body tabletop tasks." }, "name": "What is whole-body control in Gemini Robotics 2?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Gemini Robotics ER 2 is available through the Gemini API and Google AI Studio. The direct Gemini Robotics 2 VLA is offered to early-access partners, and On-Device 2 is limited to trusted testers at the time of writing." }, "name": "Is Gemini Robotics 2 available to developers?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "It achieved a reported 92% on the specific task of unscrewing a light bulb. Other multi-finger results were lower, including 36% for screwing in a bulb, 44% for tying a trash bag, 40% for closing a ziplock bag and 32% for using a dustpan." }, "name": "Did Gemini Robotics 2 achieve 92% success?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "DeepMind says On-Device 2 can adapt to new bi-arm embodiments in a few hours, typically using fewer than 200 examples. That is not a guarantee of production success and the model card notes limits on out-of-distribution tasks and high-degree-of-freedom robots." }, "name": "Can Gemini Robotics 2 adapt to a new robot with fewer than 200 examples?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "It should not be the only safety system. Google's model card excludes safety-critical use, and its safety report recommends deterministic low-level guardrails. Physical deployments still need safety-rated sensing, hardware limits, emergency stops, human authority and task-specific risk assessment." }, "name": "Is Gemini Robotics 2 safe for production robots?" }, { "@type": "Question", "acceptedAnswer": { "@type": "Answer", "text": "Start with a bounded, reversible workflow such as video-based progress checking or orchestration of existing robot APIs in simulation. This tests reasoning and tool use without immediately granting a model uncontrolled physical authority." }, "name": "What is the best first commercial use of Gemini Robotics ER 2?" } ] } ```