Hangzhou startup sells $118 AI pet collar
- Hangzhou startup Meng Xiaoyi advertised its 799 yuan ($118) PettiChat collar in May 2026, saying it uses Alibaba Cloud's Qwen model. - Meng Xiaoyi said the 27-gram collar reads more than 20 pet emotions with 94.6% to 95% accuracy and had topped 10,000 preorders. - The company has said PettiChat is due for release on May 30, with demos circulating on X and Chinese media.
Hangzhou-based startup Meng Xiaoyi is pitching a 799 yuan ($118) pet collar that it says can translate cat and dog sounds, body language and emotions into short human phrases. The device, marketed as PettiChat, uses Alibaba Cloud’s Qwen model, according to company descriptions cited by Chinese and international media. Meng Xiaoyi says the collar can identify more than 20 emotional states with 94.6% to 95% accuracy and has already drawn more than 10,000 preorders. Those figures have circulated widely on X and in follow-up coverage, but no independent validation of the claims was available in the material reviewed. ### What exactly is the startup selling? PettiChat is a lightweight collar-and-app system aimed at pet owners who want a text-style readout of what a dog or cat may be feeling. Media reports citing Meng Xiaoyi say the collar weighs about 27 grams and combines microphones with motion sensing to analyze vocalizations and movement. In demo examples described in those reports, a cat’s meow was rendered as “I wanna play,” while a dog’s bark was translated as “I’m hungry.” (indiatoday.in) Hangzhou’s municipal English-language outlet eHangzhou described the product this month as part of a cluster of local pet-tech startups. That report also repeated the 94.6% accuracy figure and said the company was working on AI tools aimed at extending pet health and communication services. ### How does Meng Xiaoyi say the collar works? (indiatoday.in) Alibaba Cloud’s Qwen model sits at the center of the company’s pitch. India Today and other outlets, citing Meng Xiaoyi, said the system was trained on millions of pet vocal samples and uses microphones, motion sensors and behavioral data to infer emotion and intent. The company has also said the device can work both ways, letting owners send simple commands back in forms the pet can recognize. (ehangzhou.gov.cn) More specialized reports said Meng Xiaoyi had collected more than 1 million cat and dog voiceprint samples and built its own pet-focused AI model on top of Qwen. That detail could not be independently confirmed from primary company documentation reviewed here, so it remains a company-linked claim repeated by secondary outlets. (indiatoday.in) ### How solid are the accuracy and preorder numbers? The most repeated numbers are 94.6% accuracy, a 799 yuan price, and more than 10,000 preorders. Those figures appear across multiple reports published on May 24 and in the earlier Hangzhou city report from May 8. The consistency suggests the numbers are coming from the company’s own marketing or distributor materials rather than from outside testing. (newsglobenow.com) No third-party benchmark, academic paper, certification filing or independent lab result was surfaced in the research reviewed. India Today explicitly noted that there were no outside verifications available at the time of writing. ### Why did this show up on X this week? (indiatoday.in) X user Pieter Levels, who posts as @levelsio, amplified the product this week in a post comparing the Hangzhou startup’s collar with AI efforts in Japan and at Rakuten. That post helped push the device into English-language social feeds, where the combination of a low price, a large claimed preorder number and the use of Alibaba’s Qwen model made it travel quickly. (indiatoday.in) The social post matched details later repeated by news outlets, including the $118 price and roughly 10,000 preorders. ### What happens next? May 30 is the date several reports gave for PettiChat’s release. Meng Xiaoyi’s next test will be whether those reported preorders convert into shipped devices and whether any outside reviewers or researchers publish evidence on the collar’s real-world accuracy after launch. (newsbytesapp.com) (indiatoday.in)