Realtime AI News
Jieyue Terminal's STEPX Neo agent phone debuts at CIIE on Qualcomm's booth
STEPX Neo, an agent phone from Jieyue Terminal, has made its first public appearance at the China International Import Expo on Qualcomm's booth. The device is pitched as an integrated software, hardware and model stack, but specifications, pricing and availability were not disclosed.
STEPX Neo, an agent phone from Jieyue Terminal, made its first public appearance at the China International Import Expo, where it is being shown on Qualcomm's booth. The pitch attached to the device is an integrated stack, with software, hardware and model treated as one system in service of agent-style experiences.
The booth tells you something, even if it does not tell you much. Appearing on Qualcomm's stand indicates a display-level association with the Qualcomm platform; the candidate material does not disclose which chip is used or what the compute budget looks like, so no specific part should be inferred.
The category label deserves unpacking. Over the past two years most phone AI shipped as single-point features, from photo cleanup to summarisation to call transcription. An agent implies tasks that run across apps in sequence, with the model reading interfaces, calling services and retrying after failure. That demands considerably more local compute, power budget and system permission than a one-shot feature.
Which is why the integrated-stack phrase keeps coming up. Pure cloud agents struggle on latency and privacy, while pure on-device agents hit power and memory ceilings. Splitting model work across device, edge and cloud is the practical compromise, and it explains why products in this category tend to be shown with a chip partner standing beside the device maker.
Choosing CIIE for a first look fits that logic. The expo reaches government, industry and the public at once, so a hardware debut there usually signals that the company wants to be seen by supply-chain partners, not only by consumers.
Confirmed detail remains thin: no price, no launch date, no display or camera specifications, and no list of third-party services the agent can call. For a first look that is normal, since a booth establishes form factor and positioning first and saves the numbers for the actual launch.
What matters next is narrower: the chip platform and on-device model size announced at launch, how reliably the agent completes real tasks, and whether the device goes to retail or starts with industry and enterprise channels. The agent-phone concept will be settled by task completion rates, not by a positioning line on a show floor.
Why it matters
Agent phones are moving from slogan to showable hardware, but the outcome will hinge on on-device compute allocation and task success rates; a first appearance is only an entry ticket.
Nearby Updates
All11/06, 17:30
Reco and Cyera rank the sensitive-data risks of AI agents
Reco and Cyera have ranked the risks AI agents create when they touch sensitive data, according to ChannelLife UK. The report does not detail the ranked items or its scoring method, but the topic points at the governance problem that follows once agents receive data permissions.
11/06, 17:54
ByteDance Seed posts Cola2, a text model tagged for diffusion and continuous latents
ByteDance's Seed team added a Cola2 entry to its official Hugging Face page, listing a text-generation pipeline with safetensors weights loaded through transformers. The tags point to diffusion, continuous latents and a base-model checkpoint, a combination rarely seen on standard language model releases.
11/06, 16:25
DeepSeek Harness discloses daily actives for the first time: over 1.5 million
DeepSeek Harness has released daily active user figures for the first time, reporting more than 1.5 million DAU, according to Yicai Global. For a team that has kept operating metrics private, the disclosure itself carries weight.
11/06, 18:37
Singapore's IMDA issues safety guidelines for AI agent use
Singapore's Infocomm Media Development Authority has published safety guidelines for the use of AI agents, according to a Lianhe Zaobao report, targeting systems that call tools and execute multi-step tasks autonomously. The guidelines' provisions, scope and whether they are mandatory still need to be confirmed from the official document.