Back to feed
News Story
量子位
1 sources

Mifeng Technology Launches 'Mifeng Pai' Embodied Data Crowdsourcing Platform, Mass-Produces 20,000 Collection Devices

Mifeng Technology launched 'Mifeng Pai' on September 23, claiming it as the world's first full-category high-quality physical AI data crowdsourcing platform that enables ordinary people to collect embodied data via hardware, app, and data engine. The company has mass-produced 20,000 MEgo collection devices and accumulated one million hours of valid data, with self-developed data engine and quality inspection system to handle complex real-world scenarios. This aims to solve the challenge of large-scale embodied AI training data collection and advance robot learning.

SynthePulse Insight · AI deep readingMembers

Mifengpai vs. Figure Index: Embodied Data Crowdsourcing and the Divide Between Self-Production and External Delivery

Version 1 · 1 source

Mifeng Technology has launched "Mifengpai," an embodied intelligence data crowdsourcing platform backed by 20,000 self-developed MEgo devices, a million hours of data, and the ManiEval grading system, attempting to break the embodied data bottleneck through socialized collection. Its most fundamental difference from Figure Index lies not in technology but in business model: one delivers externally, the other produces for itself.

  • On September 23, 2026, Mifeng Technology launched "Mifengpai," positioned as a three-in-one embodied intelligence data crowdsourcing service platform of "hardware + App + data engine."
  • Mifeng has rolled out 20,000 MEgo collection devices, accumulated a million hours of valid data covering 22 major scenario categories, and claims production capacity of tens of thousands of units per month.
  • Mifeng's ManiEval quality inspection system grades data into S, A, B, and C tiers; the team emphasizes "no filtering, only grading" with compliance and basic quality requirements as preconditions.
Open section navigationTwo Paths for Crowdsourced Data Collection: Self-Production and External Delivery

Two Paths for Crowdsourced Data Collection: Self-Production and External Delivery

On September 23, 2026, "Mifengpai," built by Mifeng Technology, made its debut, positioned as a three-in-one embodied intelligence data crowdsourcing service platform of "hardware + App + data engine." Mifeng claims this is "the world's first full-category high-quality physical AI data crowdsourcing service platform," but this "world's first" positioning lacks third-party verification and should be regarded as Mifeng's unilateral claim.

Mifengpai and Figure's Index are often compared together, but the two differ significantly in business model. Index is used by Figure for training its own models, belonging to "self-production and self-use"; Mifengpai, by contrast, faces various data procurement customers, organizes production according to customer needs, and delivers externally after processing. This difference determines that Mifengpai must handle more complex scenarios, tasks, data modalities, and quality standards, while Index only needs to serve Figure's own model training needs.

The shared judgment of both companies is: the data bottleneck of embodied intelligence cannot be solved by piling up dedicated collectors; the path of "everyone can become a robot trainer" must be made to work. Mifeng believes that the current centralized and semi-centralized data production methods cannot meet growing data demands and are relatively costly, so the industry must gradually move toward socialized collection.

20,000 Devices and a Million Hours of Data: Has Scale Been Achieved?

Mifeng's self-developed MEgo series of body-free collection terminals includes two products: MEgo View and MEgo Gripper. MEgo View uses a head-mounted panoramic camera plus a wrist close-up camera, with the head perspective recording the surrounding environment and the wrist perspective supplementing operational details; MEgo Gripper is a UMI collection gripper that is lightweight and comes standard with three-dimensional touch sensing.

As of now, Mifeng has rolled out 20,000 MEgo devices and claims this is the industry's first time such devices have been mass-produced at scale. Mifeng Chairman and CEO Yao Maoqing revealed that MEgo devices now have production capacity of tens of thousands of units per month. This means the company can rapidly increase its device inventory.

These 20,000 MEgo devices have entered real scenarios and accumulated a million hours of valid data. Mifeng claims to be the industry's first data service provider with million-hour-level body-free data available for external sale; this "industry first" claim is self-stated and lacks third-party verification.

All this data comes from real collection in open environments, covering 22 major scenario categories, with collection forms including bare hands, wrist-worn, and grippers, and data modalities covering color images, depth, inertial measurement, touch, and audio. From device volume and production capacity to data coverage, Mifeng attempts to prove that large-scale collection has been achieved, but the above data are all disclosed unilaterally by Mifeng, with no independent audit or third-party verification.

Free for now

Read the full analysis

5 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

All facts in this article come from QbitAI's report on the launch of Mifengpai. The launch date of Mifengpai, the product forms of MEgo View and MEgo Gripper, and the grading dimensions of ManiEval are facts confirmed in the report. Mifeng's claims of being "the world's first full-category high-quality physical AI data crowdsourcing service platform," 20,000 devices rolled out, a million hours of data, production capacity of tens of thousands of units per month, HandPose's seven metrics reaching SOTA, sub-centimeter-level reconstruction, MPR errors under 1 centimeter, 20,000 beta users and 13,000 tasks, some people earning over 5,000 yuan a month, more than 50 first-batch members of the Scenario Data Alliance, and serving Tencent Robotics X and Ant Lingbo are all disclosed or self-stated unilaterally by Mifeng, lacking third-party verification. Figure Index's 16 million videos, 108 countries, 264,000 downloads, 43,000 weekly active contributors, 15 million dollars in compensation, 1 billion dollar investment plan, and Helix 2.5's success rate rising from 9% to 56% all come from QbitAI's relayed account, with no first-party Figure source seen. Readers should regard the above content as source claims rather than confirmed facts.

Primary report

量子位

Primary source