hardware
Best AI Laptops
The best AI laptops and AI PCs, ranked on independent reviews and lab testing rather than launch claims: what a Copilot+ PC actually gets you, how Macs compare, how much memory local AI needs, and which machine suits which budget, operating system and kind of AI work.
Quick answer: The best AI laptop for most people is the Apple MacBook Air M5, because it is Engadget’s pick for best laptop overall, ships with 16GB of unified memory and a 512GB SSD as standard, puts a Neural Accelerator in every GPU core, and runs Apple Intelligence and small local models on a battery Apple rates at up to 18 hours, from $1,299 (Engadget, Apple). If you want Windows, the Dell XPS 14 is the pick, because Notebookcheck rated it 87% and its Intel Core Ultra Series 3 chips power Microsoft’s Copilot+ PC features, from $1,699.99. If you want to run large AI models on the laptop itself, the Microsoft Surface Laptop Ultra with Nvidia’s RTX Spark chip is the pick, with up to 128GB of memory shared between CPU and GPU. The caveat: every laptop on this page runs ChatGPT, Claude, Gemini and Copilot equally well, because those run in the cloud; “AI laptop” only matters for features that run on the device.
This page ranks laptops by how well they handle AI on the machine itself: built-in AI features, the neural processor, and the memory that decides how big a local model you can run. It does not rank the AI apps you install; for those, see best AI apps and best AI chatbots. For the AI built into handsets, see best AI phones.
Which one should you pick?
| If you want… | Buy | Why | From |
|---|---|---|---|
| The best AI laptop for most people | Apple MacBook Air M5 | Engadget’s best laptop overall, 16GB unified memory standard, all-day battery | $1,299 |
| A Windows laptop with Copilot+ features | Dell XPS 14 | Rated 87% by Notebookcheck; Intel chip that powers Copilot+ features | $1,699.99 |
| To run big AI models locally on Windows | Microsoft Surface Laptop Ultra (RTX Spark) | Up to 128GB unified memory, Nvidia Blackwell GPU | $2,599.99 |
| The cheapest laptop worth buying | Apple MacBook Neo | Best budget laptop for Engadget and PCWorld | $699 |
| The highest NPU rating on Windows | A Snapdragon X2 Copilot+ PC, such as the Surface Laptop 8 | 80 TOPS NPU; ahead in Qualcomm-commissioned lab tests | $1,599 |
| Only to use ChatGPT, Claude or Gemini | The laptop you already have | Those assistants run in the cloud | $0 |
By budget. Under $1,000, the MacBook Neo is the pick, but its 8GB of memory is a hard ceiling for local AI. Between $1,000 and $2,000, choose the MacBook Air M5 or a Copilot+ PC such as the XPS 14 depending on your operating system. Above $2,500, you are paying for local model capacity: the Surface Laptop Ultra on Windows, or a MacBook Pro with lots of memory on a Mac.
By operating system. On a Mac, every M-series model runs Apple Intelligence; the only question is memory. On Windows, buy a laptop with the Copilot+ PC label if you want Recall, Click to Do and the other NPU features; any Windows 11 laptop runs the Copilot assistant.
By on-device AI need. If you mostly use cloud chatbots, the NPU barely matters and battery, screen and keyboard should decide. If you want AI features that work offline and keep data on the device, buy an NPU laptop with at least 16GB of memory. If you want to run open-weight models locally, memory capacity is the spec that counts; see best local LLMs for which models fit which tier.
What you need to know about AI laptops
An “AI PC” is a laptop with a neural processing unit (NPU). The NPU is a block on the main chip built to run AI models at low power, so features such as live captions, webcam effects and on-device search run without draining the battery or sending data to a server. Every chip on this page has one. For how NPUs work and what their ratings mean, see what is an NPU.
A Copilot+ PC is Microsoft’s certified version. Microsoft announced the category on 20 May 2024. It requires an NPU rated at 40+ TOPS, at least 16GB of RAM and at least 256GB of storage, and lists AMD Ryzen AI 300 and 400, Intel Core Ultra 200V and 300V, and Snapdragon X series processors as qualifying (Microsoft). In return, the laptop gets Windows features that run on the NPU, including Recall (a searchable timeline of what you have seen on screen), Click to Do, live caption translation and Windows Studio Effects for the webcam (Microsoft). Microsoft said on 7 October 2026 that Copilot features “powered by hybrid intelligence are expected to begin rolling out on Copilot+ PCs in the coming months”.
Macs do not use the label but have the same hardware. Every Apple silicon Mac has a Neural Engine. From the M5 onwards Apple also builds a Neural Accelerator into each GPU core, and its headline AI claim for the M5, over four times the peak GPU AI compute of the M4, is about the GPU rather than the Neural Engine (Apple).
Memory matters more than TOPS for running models. TOPS is a peak figure each chipmaker calculates its own way. Running a language model locally means holding its weights in memory and reading them for every word it produces, so memory capacity and bandwidth decide what you can run. Macs and Nvidia’s RTX Spark laptops share one pool of memory between the CPU and GPU, which is why a 128GB machine can load models a 16GB gaming laptop cannot.
How these laptops are ranked
Each ranked laptop has more than one independent assessment: lab tests, scored reviews and buying-guide picks from outlets that tested it. Makers’ AI performance figures appear only as their claims. The full standard is on how we rank.
Best AI laptops ranked
| Rank | Laptop | Chip | AI hardware | Memory | US price from | Best for |
|---|---|---|---|---|---|---|
| 1 | Apple MacBook Air M5 | Apple M5 | 16-core Neural Engine, Neural Accelerators in each GPU core | 16GB, up to 32GB | $1,299 (13-inch), $1,499 (15-inch) | Most people |
| 2 | Dell XPS 14 | Intel Core Ultra Series 3 | Intel NPU, up to 50 TOPS on top chips | 16GB, up to 64GB | $1,699.99 | Windows and Copilot+ |
| 3 | Microsoft Surface Laptop Ultra | Nvidia RTX Spark N1X | Blackwell GPU, Nvidia NPU | 24GB to 128GB | $2,599.99 | Running large models locally |
| 4 | Apple MacBook Neo | Apple A18 Pro | 16-core Neural Engine | 8GB | $699 | Tight budgets |
The top picks, reviewed
1. Apple MacBook Air M5: best AI laptop for most people
The M5 MacBook Air is the best AI laptop for most people because Engadget rates it the best laptop overall and it ships with 16GB of unified memory as standard, the Copilot+ minimum, at a lower price than the Windows picks here. Engadget’s buying guide names it the best laptop overall (Engadget), and Digital Trends gave it an Editors’ Choice award, calling it “exceptionally good at almost everything” (Digital Trends).
The AI case. The M5 has a 16-core Neural Engine plus a Neural Accelerator in every GPU core. Every model ships with at least 16GB of unified memory and a 512GB SSD, configurable to 24GB or 32GB (Apple). That is enough for Apple Intelligence, on-device dictation and transcription, and 7–12B open-weight models at 4-bit quantisation.
Battery. Apple rates it at up to 18 hours of video streaming and 15 hours of web use (Apple). The 13-inch model lasted 13 hours 28 minutes in Digital Trends’ test.
Price. The 13-inch launched in March 2026 at $1,099. Apple raised it to $1,299 on 25 June 2026, which Bloomberg attributed to memory and storage costs (PPC Land). Engadget lists the 13-inch and 15-inch models from $1,299 and $1,499.
Limitations. It is fanless, so long local-model runs slow it down, and 16GB limits you to small models. For heavier local AI on a Mac, step up to a MacBook Pro with more memory.
2. Dell XPS 14: best Windows laptop for AI
The XPS 14 is the Windows pick because it pairs Intel chips that power Copilot+ features with an 87% rating from Notebookcheck’s lab review. Notebookcheck rates it 87% (Notebookcheck), and Engadget scored it 78 out of 100 in a review headlined “a beautiful laptop that excels at almost everything… except typing” (Engadget). Buying guides differ on the best Windows laptop: PCWorld names the Acer Aspire 16 AI its best laptop overall (PCWorld), and Engadget names the HP OmniBook Ultra its best Windows laptop.
The AI case. It runs Intel’s Core Ultra Series 3 (Panther Lake) chips, from the Core Ultra 5 325 up to the Core Ultra X9 388H. Intel rates the top Series 3 chips at 50 NPU TOPS, above Microsoft’s 40 TOPS Copilot+ bar (Intel), and at up to 180 TOPS across the whole chip (Intel). Dell says these processors “power Copilot+ PC experiences” (Dell). Intel chips run every Windows app natively, without the emulation some apps need on Arm laptops.
Battery. Engadget’s 32GB review unit lasted 10 hours 21 minutes in PCMark 10’s Modern Office battery test.
Price. Dell lists it from $1,699.99 (Dell). Engadget’s March 2026 review gave configurations from $1,450 with a Core Ultra 7 355, 16GB and 512GB, up to $2,250 with 64GB.
Limitations. The keyboard is the main complaint in Engadget’s review. A Snapdragon X2 laptop has a higher-rated NPU (80 TOPS against 50) and led Intel on CPU and AI benchmarks in a Qualcomm-commissioned test (see below).
3. Microsoft Surface Laptop Ultra: best for running AI models locally
The Surface Laptop Ultra is the best laptop for running large AI models on the machine itself, because Nvidia’s RTX Spark chip pairs a Blackwell GPU with up to 128GB of memory shared between CPU and GPU (Microsoft). Nvidia rates the chip at 1 petaflop of 4-bit (FP4) AI performance (Nvidia).
The AI case. ServeTheHome tested the 20-core, 128GB model running Qwen open-weight models of up to 35 billion parameters locally, and rated it “good, but not great” (ServeTheHome). On the same chip and memory, PCWorld’s Asus ProArt P16 finished a large gpt-oss-120b workload about 1.6 times faster than a desktop RTX 5090, with the reviewer adding that this does not make RTX Spark faster than the 5090 at AI in general (PCWorld). Microsoft’s page says the laptop “will join the Copilot+ PC category”; Nvidia publishes no NPU rating for the chip.
Price and configurations. From $2,599.99 with the 18-core chip, 24GB and 512GB; $5,899.99 for the 20-core chip with 128GB and 1TB (Notebookcheck).
Battery and build. Microsoft rates it at up to 15 hours of video playback and 12 hours of web browsing, with a 15-inch touchscreen and a weight under 4.5lb. ServeTheHome estimated 11–13 hours of light use and “closer to an hour” under heavy AI agent workloads.
Limitations. The base 24GB model holds far smaller models than the 128GB version. Reviews of the RTX Spark platform are mixed: ServeTheHome’s verdict is “good, but not great”, and Paul Thurrott called RTX Spark “crushingly disappointing” after using three RTX Spark laptops (Thurrott).
4. Apple MacBook Neo: best budget pick
The MacBook Neo is the best budget laptop for everyday AI use because it gets you macOS, a 16-core Neural Engine and strong battery life for $699. Engadget and PCWorld both name it the best budget laptop (Engadget, PCWorld); Engadget scored it 90 out of 100 at launch (Engadget) and Notebookcheck 86% (Notebookcheck).
The AI case. It uses the A18 Pro, the chip from the iPhone 16 Pro, with a 16-core Neural Engine and 8GB of memory in every model, with 256GB or 512GB of storage (Apple). That runs Apple Intelligence and small 3–4B local models.
Battery. Apple rates it at up to 16 hours of video streaming. It lasted 12 hours 15 minutes in Engadget’s video test.
Price. It launched in March 2026 at $599; Apple has since raised it to $699 (Engadget).
Limitations. The 8GB of memory cannot be upgraded: Engadget notes there is “no way to add more RAM”, and Notebookcheck lists the soldered memory as a drawback (it names the missing keyboard backlight as the biggest issue). For local AI, 8GB means small models only. If AI on the device matters to you, the MacBook Air is worth the extra.
The AI chips, compared
| Chip | Found in | NPU rating (maker’s figure) | Copilot+ | Note |
|---|---|---|---|---|
| Qualcomm Snapdragon X2 Elite and X2 Plus | Surface Laptop 8 and others | 80 TOPS (INT8) | Yes | Ahead of Intel and AMD on CPU and AI tests in a Qualcomm-commissioned report (see below) |
| Intel Core Ultra Series 3 (Panther Lake) | Dell XPS 14, Acer Aspire 16 AI and others | Up to 50 TOPS (top chips) | Yes | Runs Windows apps without emulation |
| AMD Ryzen AI 300 and 400 | Framework Laptop 13 and others | 50 to 60 TOPS (Ryzen AI 400) | Yes | On Microsoft’s Copilot+ processor list |
| Nvidia RTX Spark N1X | Surface Laptop Ultra, Asus ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra, Lenovo Yoga Pro 9n, MSI Prestige N16 | Not published | Microsoft says the Surface Laptop Ultra “will join” | 1 petaflop FP4 for the whole chip; up to 128GB unified memory |
| Apple M5 | MacBook Air | Not published | Not applicable | Neural Accelerators in every GPU core |
| Apple M6 | 14-inch MacBook Pro | Not published | Not applicable | Apple claims a Neural Engine up to 1.9 times faster than M5 |
| Apple A18 Pro | MacBook Neo | Not published | Not applicable | iPhone-class chip, 8GB memory |
Sources: Qualcomm, Intel, AMD, Microsoft, Nvidia, 9to5Mac.
Qualcomm leads on NPU rating, and in its own commissioned testing. A Signal65 report commissioned by Qualcomm, published on 13 August 2026, found the Snapdragon X2 Elite Extreme up to 53% faster than Intel’s Core Ultra X9 388H and up to 83% faster than AMD’s Ryzen AI 9 465 in Geekbench 7, and “nearly 2x” the Intel system’s score in Procyon AI Computer Vision. The Snapdragon laptop ran on battery while the Intel laptop was plugged in (Signal65). Qualcomm paid for the study, so it is the vendor’s claim rather than independent testing.
TOPS tells you whether a laptop qualifies, not how fast it is. Each vendor picks the precision and workload behind its figure, Apple publishes no figure at all, and Nvidia rates RTX Spark’s whole chip in petaflops of 4-bit maths, which cannot be compared with an NPU’s 8-bit TOPS.
How much memory you need for AI
| Memory | What runs locally | Example laptops |
|---|---|---|
| 8GB | Built-in AI features; 3–4B models only | MacBook Neo |
| 16GB | Copilot+ minimum; 7–12B models at 4-bit | MacBook Air M5, most Copilot+ PCs |
| 32GB | 20–35B models at 4-bit, such as gpt-oss-20b or Qwen 3.6 27B | XPS 14 higher configurations, MacBook Pro M6 |
| 64GB | 70B-class models at 4-bit | XPS 14 top configuration, MacBook Pro M5 Max configurations |
| 128GB | 120B-class models such as gpt-oss-120b | Surface Laptop Ultra and Asus ProArt P16 top configurations |
Model sizes match the hardware tiers on the best local LLMs page. On Windows laptops with a separate graphics card, what counts is the card’s own video memory, not system RAM, which is why a gaming laptop with an 8GB GPU runs smaller models than a 32GB Mac.
Other laptops worth knowing (not ranked)
These are machines you can buy that either launched too recently for a full review record, cost more than most readers will pay, or did well with one reviewer but not yet several.
Windows laptops
| Laptop | Chip | Price from | Why it’s interesting |
|---|---|---|---|
| Microsoft Surface Laptop 8 | Snapdragon X2 Plus or X2 Elite | $1,599 | 80 TOPS NPU, 13.8-inch 120Hz touchscreen; Notebookcheck calls it a “very good everyday laptop” but “too expensive” against rivals such as the Asus Zenbook A14 (Notebookcheck) |
| Acer Aspire 16 AI | Intel Core Ultra 7 355 | $1,279.99 (PCWorld’s price when reviewed) | PCWorld’s best laptop overall pick (PCWorld) |
| Framework Laptop 13 | AMD Ryzen AI 300 series | Varies by build | Repairable and upgradeable; PCWorld’s pick for upgrading |
For heavy local AI
Apple MacBook Pro 14-inch (M6). Announced on 27 October 2026, with first orders arriving on 6 November, at $1,999 with 16GB and 1TB; a 32GB model costs $2,399. Apple says the M6’s dual 16-core Neural Engine is up to 1.9 times faster than the M5’s, and that time to first token in large language model prompt processing is up to 1.3 times faster than the M5 14-inch MacBook Pro (9to5Mac). Those are Apple’s figures. The M5 Pro and M5 Max MacBook Pros, which take far more memory, stay on sale alongside it (PPC Land).
Asus ProArt P16 (RTX Spark). The creator version of RTX Spark, $6,999 with 128GB and a 4K OLED screen. In PCWorld’s test it finished a large gpt-oss-120b workload about 1.6 times faster than a desktop RTX 5090, but its reviewer’s verdict for most buyers was “Hell no”, citing the price, fans that “get extremely loud” and a battery that “craters during sustained AI inference” (PCWorld).
HP OmniBook Ultra 16 and Lenovo Yoga Pro 9n. Two more RTX Spark laptops from Nvidia’s launch partners (Nvidia). Paul Thurrott used both, alongside the ProArt P16, for his RTX Spark review.
AI tablets
If you want a tablet rather than a laptop, the same rules apply. The Microsoft Surface Pro with a Snapdragon X chip is a Copilot+ PC in tablet form and PCWorld’s pick for the best lightweight laptop (PCWorld). Apple’s iPad Pro and iPad Air run Apple Intelligence on M-series chips; Apple raised their US starting prices to $1,199 and $749 on 25 June 2026 (PPC Land).
What an AI laptop does, and doesn’t, change
What runs on the laptop. Copilot+ features (Recall, Click to Do, live caption translation, Studio Effects), Apple Intelligence’s on-device models, webcam background blur and eye contact, offline dictation, photo search, and any open-weight model you run yourself through tools such as LM Studio or Ollama.
What runs in the cloud regardless. ChatGPT, Claude, Gemini, the Copilot assistant, image and video generators, and every frontier model in the best AI models ranking. A five-year-old laptop runs these as well as a new one.
The practical test. Name the two or three AI things you will actually do. If they are all chatbots, buy for battery, screen and keyboard. If one of them is a Copilot+ feature, buy a Copilot+ PC. If one is running models locally, buy memory first and everything else second.
Frequently asked questions
What is the best AI laptop?
The Apple MacBook Air M5 is the best AI laptop for most people. It is Engadget’s pick for best laptop overall, comes with 16GB of unified memory and a 512GB SSD as standard, has a Neural Accelerator in every GPU core, and Apple rates its battery at up to 18 hours, from $1,299. On Windows, the Dell XPS 14 is the pick, rated 87% by Notebookcheck, from $1,699.99. For running large AI models locally, the Microsoft Surface Laptop Ultra with Nvidia RTX Spark is the pick, with up to 128GB of memory.
What is an AI PC?
An AI PC is a laptop or desktop with a neural processing unit (NPU), a part of the chip built to run AI models at low power. The NPU handles features such as live captions, webcam effects and on-device search without sending data to a server or draining the battery. Every laptop chip on this page includes one. Microsoft’s certified version of an AI PC is called a Copilot+ PC.
What is a Copilot+ PC?
A Copilot+ PC is a Windows 11 laptop or desktop that meets Microsoft’s hardware bar for on-device AI: an NPU rated at 40+ TOPS, at least 16GB of RAM and at least 256GB of storage. Microsoft announced the category on 20 May 2024. Copilot+ PCs get Windows features that run on the NPU, including Recall, Click to Do, live caption translation and Windows Studio Effects. Microsoft lists AMD Ryzen AI 300 and 400, Intel Core Ultra 200V and 300V, and Snapdragon X series processors as qualifying.
Do I need an AI laptop to use ChatGPT or Copilot?
No. ChatGPT, Claude, Gemini and the Microsoft Copilot assistant run on servers in the cloud, so any laptop with a web browser runs them equally well. An AI laptop only adds features that run on the device itself, such as Windows Recall, Apple Intelligence’s on-device models, offline transcription and open-weight models you run locally.
Is a MacBook an AI laptop?
Yes. Every Apple silicon Mac has a Neural Engine and runs Apple Intelligence, and from the M5 onwards Apple adds a Neural Accelerator to every GPU core. Macs do not carry Microsoft’s Copilot+ PC label because that is a Windows certification. Their unified memory, shared between CPU and GPU, also makes high-memory MacBook Pros among the best laptops for running large AI models locally.
How much RAM do I need for AI on a laptop?
16GB is the practical minimum: it is Microsoft’s Copilot+ requirement and enough to run 7–12B open-weight models at 4-bit quantisation. 32GB runs 20–35B models such as gpt-oss-20b or Qwen 3.6 27B. Running 70B-class models needs 64GB, and 120B-class models such as gpt-oss-120b need around 128GB of unified memory. An 8GB laptop such as the MacBook Neo handles built-in AI features and only very small local models.
How many TOPS do I need?
You need an NPU rated at 40+ TOPS for a Windows laptop to qualify as a Copilot+ PC. Current laptop chips are rated at up to 50 TOPS (Intel Core Ultra Series 3, top chips), 50 to 60 TOPS (AMD Ryzen AI 400) and 80 TOPS (Qualcomm Snapdragon X2). Beyond the Copilot+ bar, extra TOPS rarely changes what you can do, because each vendor measures TOPS its own way and memory decides which models you can run.
Is Snapdragon, Intel or AMD better for a Copilot+ PC?
Snapdragon X2 has the highest NPU rating at 80 TOPS, and a Qualcomm-commissioned Signal65 report found the Snapdragon X2 Elite Extreme ahead of Intel’s Core Ultra X9 388H and AMD’s Ryzen AI 9 465 on CPU and AI benchmarks. Intel’s Core Ultra Series 3 runs every Windows app natively, without the emulation some apps need on Arm, and powers the Dell XPS 14. AMD Ryzen AI 300 and 400 chips are also on Microsoft’s Copilot+ list. For most buyers, the specific laptop’s reviews matter more than the chip brand.
What is the best laptop for running AI models locally?
On Windows, the Microsoft Surface Laptop Ultra with Nvidia RTX Spark, which comes with up to 128GB of memory shared between CPU and GPU; on the same chip and memory, PCWorld ran a gpt-oss-120b workload on the Asus ProArt P16. On a Mac, a MacBook Pro with as much unified memory as you can afford is the equivalent. Memory capacity, not NPU TOPS, decides which models fit.
What is the best cheap AI laptop?
The Apple MacBook Neo, from $699, is the best budget laptop according to Engadget and PCWorld. Its 8GB of memory, which cannot be upgraded, limits on-device AI to built-in features and very small models. If you need Windows, the Acer Aspire 16 AI is PCWorld’s best laptop overall pick, at $1,279.99 when PCWorld reviewed it. Spending up to the $1,299 MacBook Air M5 doubles the memory, which matters more for AI than any other upgrade.
Is the RTX Spark Surface Laptop Ultra worth it?
Only if you plan to run large AI models on the laptop itself. Its 128GB configuration, at $5,899.99, holds models of a size that 16GB and 32GB laptops cannot load. ServeTheHome rated it “good, but not great”, estimated battery life at closer to an hour under heavy AI agent workloads, and Paul Thurrott called the RTX Spark platform “crushingly disappointing”. Microsoft’s page says the laptop “will join the Copilot+ PC category”.