AISpeech has introduced a new dedicated voice recorder in China, the TalkNote Flagship AI Recording Card, priced at 1,299 yuan (around $192). The device is designed to handle audio recording, transcription, and cloud syncing in a form factor small enough to fit inside a wallet or a standard office badge holder.

AISpeech TalkNote Specifications

Physically, the TalkNote measures 86 x 55 x 4.5 mm and weighs 43 grams, giving it the approximate footprint of a credit card. It features a grey finish and includes a small 0.95-inch AMOLED display. The screen is straightforward, used primarily for checking battery life and current recording status rather than navigating complex menus.

For audio capture, the device relies on a five-microphone array consisting of four omnidirectional microphones and a single bone-conduction microphone. According to AISpeech, this hardware combination allows for clear voice pickup within a five-meter radius, with a maximum effective range of eight meters. It also utilizes onboard noise reduction to filter out ambient background sound in busy environments.

The device comes with built-in 4G, alongside standard Wi-Fi and Bluetooth. AISpeech includes lifetime free 4G data specifically for cloud synchronization. This allows the device to automatically upload audio files to the cloud as soon as a recording session ends, meaning users don’t have to manually transfer files to a computer or phone via the included USB-C port. The device is powered by a 750mAh battery, which the company says provides up to 45 hours of continuous recording and 60 days of standby time.

On the software side, the TalkNote is built around automated transcription. It supports 25 foreign languages and 57 Chinese dialects, and it can automatically switch between Chinese and English during a conversation. AISpeech claims a transcription accuracy rate of up to 98% under supported conditions. Once the audio is uploaded, the companion software can generate meeting summaries, identify different speakers, and extract action items.

The device comes with 64GB of internal storage and includes a three-month subscription to AISpeech’s OfficeClaw platform, which manages the AI-generated notes and tasks.

In related news, Lenovo has recently introduced a new AI-powered gadget for kids featuring a flip camera, GPS tracking, and an SOS button.

For more daily updates, please visit our News Section.

Stay ahead in tech! Join our Telegram community and sign up for our daily newsletter of top stories!

(JD)

The post AISpeech launches credit card-sized AI voice recorder with built-in 4G, AI summaries & automatic cloud uploads appeared first on Gizmochina.

In a pretty big surprise for the AI world, China’s Zhipu AI’s latest model, GLM-5.2, just grabbed the top spot in one of the most respected creative coding benchmarks out there.

On June 19, 2026, Design Arena announced on X that GLM-5.2 has taken the #1 position in their single-round HTML web design leaderboard (non-agent category). The platform, basically the world’s first big crowdsourced benchmark where real people blindly vote on AI-generated designs, says it beat out Claude Fable 5, along with the Opus 4.6 and 4.7 versions. That’s a solid five-place jump from its predecessor, GLM-5.1, landing it with an Elo score around 1360 in the Code Categories arena.

What makes this win stand out is how the model actually designs. It’s apparently excellent at clean layouts, pulling in images from CDNs, typography, visual hierarchy, and those nice subtle animations that make a site feel alive. It also plays really nicely with popular libraries like Chart.js and Three.js. According to the results, it boosted its win rate by about 6 percentage points this time around.

One interesting detail: GLM-5.2 leans heavily on Tailwind CSS (showing up in 91% of its designs) and Font Awesome (51%). By comparison, Fable 5 only used Tailwind in about 57% of its sessions, which might explain some of the practical gaps people noticed.

Beyond the design quality, the pricing is a breath of fresh air. GLM-5.2’s API runs around $1.40 per million input tokens and $4.40 for output, way more affordable than Fable 5’s $10/$50. Plus, it’s an open-weights model under the MIT license with a solid 1M-token context window, which means teams can run it locally and tackle bigger, longer projects without being locked into a proprietary service.

Design Arena’s approach gives these results real weight: millions of votes from actual creators focused on real aesthetics and usability, not just synthetic tests. This isn’t just a win for web design; it puts GLM-5.2 among the strongest open models across broader coding benchmarks too.

For the AI community, this feels like another sign that Chinese labs are moving fast in specialized creative areas. While Claude Fable 5 is a strong player, GLM-5.2’s mix of strong performance, low cost, and openness could make it a go-to choice for a lot of web development teams going forward.

Don’t miss a thing! Join our Telegram community for instant updates and grab our free daily newsletter for the best tech stories!

For more daily updates, please visit our News Section.

(Source: Design Arena)

The post China’s GLM-5.2 beats Claude Fable 5 in web design, claims No. 1 spot appeared first on Gizmochina.