Lenovo unveils Qira, a cross-device AI assistant designed to act for users

Lead

At CES, Lenovo unveiled Qira, a system-level AI assistant intended to run across Lenovo laptops and Motorola phones, announced during a keynote at The Sphere in Las Vegas. The company pitched Qira as a cross-device intelligence that combines on-device and cloud models to provide continuity, context and the ability to perform tasks on users’ behalf. Jeff Snow, Lenovo’s head of AI product, described Qira as the firm’s most ambitious AI effort, born from a company-wide reorganization of AI teams less than a year ago. The release signals a strategic push by the world’s largest PC shipper to make AI a built-in differentiator rather than a bolt-on feature.

Key Takeaways

• Qira was introduced at CES during an event at The Sphere in Las Vegas and is intended to operate across Lenovo laptops and Motorola phones.
• Lenovo consolidated AI teams under a single software-focused group less than a year ago to accelerate cross-device AI development.
• The system is modular, combining local on-device models with cloud models accessed via Microsoft/Azure and OpenAI, plus integrations with Stability AI, Notion, and Perplexity.
• Lenovo retains distribution reach: it ships tens of millions of devices annually, giving it direct influence over mainstream AI exposure.
• Privacy features are explicit: Qira uses opt-in memory, visible recording indicators, and user controls to avoid silent collection.
• Performance favors higher-end hardware; Lenovo is optimizing local models to run acceptably on machines with around 16 GB of RAM.
• Qira aims to improve device retention and differentiate Lenovo as hardware specs converge across the market.
• Lenovo framed Qira as focusing on continuity, context and acting on the device rather than competing with prompt-driven chatbots.

Background

Lenovo has long been organized around hardware SKUs, supply chains and global manufacturing scale. That orientation made it one of the most effective companies at getting devices into millions of hands each year, a position that now intersects with the rapid rise of generative AI. In response, Lenovo moved AI specialists out of isolated product units for PCs, tablets and phones into a centralized software group less than a year ago, signaling a shift from hardware-first to AI-infused product thinking.

Earlier experiments informed the new approach. Snow noted lessons from Moto AI, an assistant that generated strong initial curiosity—more than half of Motorola users tried it—but suffered weak long-term retention because it resembled prompt-based chat features available elsewhere. Separately, the industry response to Microsoft’s Recall showed how privacy missteps can generate backlash, prompting OEMs to design more transparent memory and control systems from the outset.

Main Event

At the CES showcase, Lenovo presented Qira as a system-level assistant that persists across devices and sessions. Rather than betting on a single flagship model, the company described a modular architecture that routes tasks to the best available engine: local models for latency and privacy-sensitive tasks, and cloud models for heavier reasoning or specialized capabilities. Snow demonstrated using an on-device model during a flight to refine remarks for CES meetings, emphasizing that local inference was practical for certain workflows.

Under the hood, Lenovo said Qira will mix on-device models with cloud backends anchored by Microsoft and OpenAI via Azure, and will also include Stability AI’s diffusion models for creative outputs and app-specific partners such as Notion and Perplexity. The approach is intentionally agnostic: Lenovo argued different tasks require different tradeoffs across performance, cost and quality, so optionality prevents vendor lock-in and preserves flexibility as models evolve.

Privacy and transparency were front and center in the presentation. Qira’s memory is opt-in, with persistent indicators when recording or ingesting context, and user-facing controls to review or delete stored data. Lenovo framed these design choices as a direct response to consumer concerns and to the negative reactions that followed earlier industry experiments with always-on capture.

Analysis & Implications

Lenovo’s scale gives it a potentially outsized role in how consumer AI is experienced. Because the company ships tens of millions of devices per year, decisions about default assistants, bundled services and cross-device integrations can shape motion toward particular ecosystems and usage patterns. That distribution advantage is why Lenovo emphasizes modularity: it can offer choice while retaining access to large swaths of users.

The modular architecture also reflects an economic reality: running large models exclusively in the cloud is expensive, and on-device inference places pressure on memory and compute. Analysts already warn that rising memory prices and AI-driven component demand may push PC costs higher. By optimizing local models to smaller footprints—targeting reasonable operation on machines with about 16 GB of RAM—Lenovo is trying to balance cost, privacy and performance for a broad installed base.

Strategically, Qira is both a retention lever and a hedge against commoditization. In markets where hardware specifications are increasingly similar, software-level differentiation and seamless device continuity become key selling points. If Qira delivers genuine productivity gains tied to Lenovo hardware, it could slow churn and justify premium positioning; if not, it risks becoming another preinstalled layer users ignore.

Finally, Lenovo’s insistence on partnerships rather than exclusivity could influence how other OEMs approach AI. Major labs and cloud vendors often seek tight integrations; Lenovo’s multi-vendor stance encourages an ecosystem where performance and cost tradeoffs determine routing rather than single-vendor lock-in. That posture may complicate engineering but could benefit users through competition and functionality diversity.

Comparison & Data

The table highlights why Lenovo is pursuing a hybrid approach: on-device models reduce latency and improve privacy but are constrained by device memory and CPU, while cloud models offer higher capability at the expense of latency and recurring costs. Lenovo’s engineering goal is to push local model efficiency down to typical consumer RAM levels without eroding the perceived experience.

Reactions & Quotes

Lenovo executives framed Qira as a deliberate pivot toward integrated AI that respects user control. The company emphasized learnings from prior efforts and industry missteps to explain the assistant’s safeguards and modular design.

We designed Qira to provide continuity and take actions on the device while keeping memory and recording under user control.

Jeff Snow, Lenovo head of AI product (paraphrased)

Outside observers noted the strategic logic and the engineering challenge. An independent industry analyst said Lenovo’s distribution could normalize AI assistants if Qira finds genuine daily utility, but warned that execution and user trust will determine adoption.

Lenovo’s scale could accelerate mainstream assistant use, but the product must deliver clear utility and trustworthy controls to stick.

Industry analyst (anonymous)

Unconfirmed

• Exact timelines for Qira rolling out to all Lenovo laptops and Motorola phones remain unannounced and are subject to internal scheduling.
• Claims about Qira performing acceptably on 16 GB machines are targets from Lenovo engineering; broad performance benchmarks across models are not yet independently verified.
• The long-term effect on PC pricing due to AI-related memory demand is forecasted by analysts but remains uncertain and dependent on supply dynamics.

Bottom Line

Qira represents a notable step by a major OEM to embed AI across devices with an emphasis on modularity, privacy controls and cross-device continuity. Lenovo is leveraging distribution scale while attempting to avoid single-vendor lock-in by combining local inference with cloud partners. Whether Qira becomes a meaningful retention and differentiation tool will depend on execution: performance on mid-range hardware, the clarity and usability of privacy controls, and the assistant’s ability to perform unique actions rather than mimic chatbots.

For customers and competitors alike, the key signals to watch are rollout scope, demonstrable day-to-day productivity gains, and whether Lenovo’s hybrid model balance helps contain costs without undermining user experience. If Qira meets those tests, it could accelerate the presence of persistent AI assistants in mainstream consumer devices; if it falls short, it will join earlier attempts that generated initial interest but weak retention.

Sources

• The Verge — technology news reporting on Lenovo’s CES announcement
• Lenovo Newsroom — official company announcements and event coverage