LG is trying something a little clever at CES this year: stitch a genuinely useful on-device AI into an ultraportable that still promises days of battery life. The company’s new gram Pro AI machines — the 16-inch model in particular — pair an upgraded local language model called Exaone 3.5 with a 77Wh cell that LG says can power the laptop for up to 27 hours on a single charge. The framing is deliberate: this isn’t just about the novelty of having an assistant on your machine, it’s about making that assistant reliable when you’re offline and unobtrusive when you’re not.

What LG showed and what it promises are two related but separate propositions. On the software side, Exaone 3.5 is central: LG AI Research built the model so that routine productivity tasks — summarising documents, translating passages, auto-drafting contextual replies and smart searching across your files — can be handled locally, without routing sensitive material to cloud services. That’s the argument for privacy and predictable latency: if the heavy lifting happens on the laptop, flaky Wi-Fi is less of a problem. LG has also bundled a feature called My Archive that indexes a user’s files into a private, searchable knowledge base so the local model can answer questions about what’s already on the machine. On select SKUs, LG pairs this on-device layer with cloud services like Microsoft Copilot+ PC and its own gram Chat, effectively offering a “dual AI” mode that can swap between privacy and extra cloud horsepower depending on the task. That architecture is the pitch: local-first AI for everyday work, cloud for larger models or heavier compute.

The battery claims are the other hook, and they’re striking because long life and local AI usually pull in opposite directions. The 16-inch gram Pro’s 77Wh battery is the headline figure — LG’s own testing gives it up to 27 hours of video playback or “general use” — and there’s a fast-charge promise of more than nine hours of runtime after a 30-minute top-up. Those numbers will almost certainly compress once you start pushing Exaone, multi-tasking with higher refresh displays, or running GPU-heavy creative apps, but the marketing point is clear: this is a machine targeted at people who move between flights, meetings and cafes and don’t want to carry power bricks.

If the battery sells the gram Pro AI to travellers, the chassis should quiet the nagging worry that “light” usually equals “fragile.” LG says it’s abandoning older plastic shells for an aerospace-inspired alloy it calls Aerominum — a magnesium-aluminium composite intended to marry the lightness of magnesium with aluminium’s scratch resistance. The 16-inch gram Pro is listed at roughly 1.199kg, and LG claims improved scratch resistance and passage of seven U.S. military reliability tests, including shock, dust, vibration and extremes of temperature. Taken together, these details shape the device as a premium, road-worthy ultraportable rather than a fragile proof-of-concept.

Under the skin, LG isn’t forcing a one-size-fits-all approach to silicon. The 2026 gram family will be offered across both Intel and AMD platforms: Core Ultra processors appear in some models while others ship with AMD’s Ryzen AI 400 series. That gives LG flexibility to tune a given SKU for raw compute, integrated graphics, or sheer battery efficiency. The company also showed multiple form factors in its CES slate — a convertible gram Pro 360 AI and a Grambook AI variant among them — and said the domestic rollout begins in early January; CES will be the global splash that positions the Exaone-equipped gram Pro AI as the halo model for the range.

Beyond the spec sheet, what matters is whether the idea of a local, file-aware assistant actually changes people’s behaviour. Laptop makers have spent the last decade squeezing weight and battery from the same tin; adding a resident language model raises fresh questions about thermal headroom, noise, and real-world battery drain. If Exaone 3.5 lives up to its billing — fast and private for everyday tasks, but smart enough to make Apple-style continuity feel unnecessary — the gram Pro AI could be one of the more convincing “AI PC” demos at CES. For a buyer who values mobility and privacy over raw GPU muscle, the proposition is seductive: an assistant that sits on your machine, knows your files, and doesn’t demand a constant cloud connection.

Sceptics will ask sensible follow-ups. How big an impact does Exaone have on battery life when you’re using it all day? How seamless is the handoff between local intelligence and cloud services such as Copilot+ PC? And crucially, how much storage and memory will LG reserve for that on-device intelligence before it starts to feel like a permanent tenant in your SSD? LG has sketched the architecture, but the answers will arrive when reviewers get hands-on time and independent testing begins to measure throughput, thermals and battery under mixed, AI-heavy workloads.

For now, LG’s move is notable for two converging trends: hardware makers are treating on-device language models as core platform features, not optional extras, and they’re trying to deliver that capability without forcing users to trade away the battery life that defined the gram line. If LG can genuinely combine a multi-day battery with a helpful, private assistant in a 1.2-kg shell, rivals will face a simple sales question from buyers: why not the same? The real test won’t be LG’s slide deck at CES, but whether those claims survive sustained, real-world use.