PokeBot said on 30 October it is “introducing RoboAPI: a publicly accessible, ready-to-use API for robots in the physical world.” It says “During the technical preview, RoboAPI will be available free of charge to everyone”.
On PokeBot’s own benchmark, its Poke-Aurora model scored 86.0% on transfer and organization tasks against 30.9% for ฯ0.5, about 2.8 times the success rate, by our arithmetic. PokeBot built the benchmark.
Business Pill · PHYSICAL AI EXPLAINED
A short explainer on physical AI, models that perceive and act in the real world. It teaches the general idea only and says nothing about PokeBot or this launch.
The key insight: PokeBot has opened RoboAPI free during a technical preview for non-commercial use; on its own benchmark, Poke-Aurora’s 86.0% transfer success is about 2.8 times ฯ0.5’s 30.9%, by our arithmetic.
PokeBot says RoboAPI, “Taking robots’ observations and state as input”, “follows user instructions and predicts the next action chunk in near real time.” Its stated goal is “zero-shot generalization: enabling robots to follow user instructions without additional data collection or model training.”
It writes: “We trained Poke-Aurora, our first world-action model, to make zero-shot generalization readily accessible.” It adds that “mounting an action camera on the wrist of a robot arm is enough for RoboAPI to control it for task execution, provided the arm has sufficiently capable inverse kinematics (IK) solving.”
PokeBot reports that the model “transferred successfully to entirely unfamiliar robot configurations, including substantially different ones (e.g., moving from a UR7e to a YAM Standard).” It says switching fisheye camera brands “had almost no effect on performance.”
Success rates by manipulation goal on PokeBot’s own benchmark, as published on pokebot.com/ai on 30 October 2026: Poke-Aurora, Astra, pi0.5 and MolmoAct2. The benchmark was built and run by PokeBot. 2.8x (86.0 / 30.9) is our arithmetic.
The Benchmark PokeBot Published
PokeBot’s page lists success rates in three categories. In transfer and organization, Poke-Aurora scored 86.0%, Astra 71.3%, ฯ0.5 30.9% and MolmoAct2 26.9%. In pose and shape alignment the figures were 54.1%, 36.3%, 14.8% and 7.9%; in environment interaction, 87.7%, 67.9%, 32.9% and 23.6%.
PokeBot cautions: “While Poke-Aurora leads across the benchmark we built, we are well aware of the complexity of the real world and cannot guarantee that your experience with RoboAPI will exactly match the statistical results.” We covered Physical Intelligence’s ฯ API, which offers fine-tuning, earlier this month.
Free, but Not Commercial
A footnote on the page says “RoboAPI is currently a non-commercial offering and does not provide products or services to for-profit third parties.” PokeBot also notes that “Instructions in English generally yield better performance.” and that “RoboAPI’s safety gate rejects instructions that could potentially harm people or the surrounding environment.”
Gasgoo AutoNews reported that PokeBot, founded in April 2026, closed a nine-figure Pre-A round, after an angel round led by Yunqi Partners in late April.
The Structural Read
As we read PokeBot’s page, the product claim is zero-shot: robot arms following instructions without new data collection or training.
As we read the hardware note, the entry requirement is a wrist-mounted action camera and an arm with capable inverse kinematics.
As we read the benchmark, PokeBot built and ran it, and PokeBot itself says it cannot guarantee real-world results will match.
PokeBot, RoboAPI announcement, 30 October 2026
“While Poke-Aurora leads across the benchmark we built, we are well aware of the complexity of the real world and cannot guarantee that your experience with RoboAPI will exactly match the statistical results.”
Three Implications
FOR ROBOT BUILDERS PokeBot says RoboAPI is free during the technical preview and non-commercial for now.
FOR MODEL RIVALS PokeBot’s own benchmark lists Astra, ฯ0.5 and MolmoAct2 below Poke-Aurora in all three categories.
FOR POKEBOT Gasgoo AutoNews reported a nine-figure Pre-A round for the company, founded in April 2026.
The framework says “It is a layered industrial stack — and at any given moment, the binding constraint sits at a different layer than it did six months ago.”
As we read it, the framework describes AI as a layered stack; PokeBot’s RoboAPI sits at the physical-action end, and the framework predates this launch and does not discuss PokeBot.
What Is Not Established
The benchmark is PokeBot’s own and has not been independently replicated in what we read. PokeBot does not publish pricing after the technical preview, and Gasgoo does not give the exact size of the Pre-A round.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
PokeBot has opened RoboAPI free during a technical preview for non-commercial use. On its own benchmark, Poke-Aurora’s 86.0% transfer success is about 2.8 times ฯ0.5’s 30.9%, by our arithmetic.
A note on sourcing. Product details, quotes and benchmark figures come from PokeBot’s 30 October RoboAPI page; the benchmark is PokeBot’s own. Funding details come from Gasgoo AutoNews. The 2.8x ratio is our arithmetic. We did not test RoboAPI or contact PokeBot.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
