Dell XPS 13 vs MacBook Neo: Which Is Better for Developers?
Hands-on dev benchmarks compare Intel's Core Ultra XPS 13 and Apple's A18 Pro MacBook Neo on compiling, local LLM inference, and battery.
What happens when a Core Ultra XPS 13 and an A18 Pro MacBook cost the same?
Dell’s new XPS 13 (Intel’s Core i5 “Wildcat”) now sells for the same price as Apple’s entry level MacBook, nicknamed the MacBook Neo, which runs the A18 Pro chip. That price parity makes a direct comparison worth running, especially for developers who care less about chassis design and more about compile times, local AI inference, and how a laptop behaves once it’s unplugged. Independent testing of both machines on real dev workloads shows a split result: Intel wins on raw compute and compiling, Apple wins on sustained performance and battery efficiency.
TL;DR
- The Intel Core Ultra XPS 13 finished a multi-core Python Mandelbrot benchmark in about 65 seconds on average versus roughly 87 to 92 seconds for the A18 Pro MacBook Neo, a win of around 21 seconds.
- Surprisingly, the XPS held that lead even on battery power, historically a weak spot for Intel laptops that needed wall power to hit peak performance.
- Under sustained load (five back-to-back benchmark runs), the fanless MacBook Neo throttled about 7%, while the actively cooled XPS stayed faster overall but ran roughly twice as hot in relative terms and showed more run-to-run variance.
- Compiling llama.cpp from source took 3 minutes 28 seconds on the XPS versus 4 minutes on the Neo, though the Mac build also compiled Apple’s Metal GPU backend as part of that process.
- For local LLM inference with a Qwen3 4B model at 4-bit quantization, the XPS edged ahead on prompt processing (156 vs 142 tokens/sec) while the Neo won token generation (17.9 vs 15.7 tokens/sec).
- Neither machine has enough usable GPU memory to run an 8B parameter model without spilling into system RAM, and the XPS dropped to 0.23 tokens per second in that scenario, effectively unusable, while the Neo still managed a workable 7.1 tokens per second.
- On a repeated battery test combining Python compiling and LLM inference, the Neo used about 13% of its battery while the XPS burned through 21% for the same finished work, roughly 60% more battery for equal output.
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
How do the two laptops compare on hardware and build?
Both machines are nearly the same size and weight class: the XPS 13 weighs about 1.56 kg and the MacBook Neo about 1.24 kg. Both open with one finger and both feel solidly built for their price point. Configuration-wise, the tested units matched closely: 8GB of RAM in each, but the XPS shipped with a 512GB SSD against the Neo’s 256GB, and the XPS drive is user-replaceable while the Mac’s storage is soldered and fixed for life.
Storage speed is not close. The XPS hit sequential read/write speeds of 6,860 and 4,873 MB/s, while the Neo measured 1,735 and 1,500 MB/s, a gap attributed partly to smaller SSD chips in the lower-capacity Mac configuration being inherently slower. Random read/write performance (the pattern that matters most during compiling, where many small files get touched) was mixed: the Neo’s random read edged out the XPS (549 vs 237), but the XPS crushed it on random write (635 vs 52).
Display and input differences matter for daily coding. The XPS offers a 120Hz touchscreen with a matte finish, noticeably smoother for scrolling code, plus a backlit keyboard with adjustable brightness levels. The Neo is capped at 60Hz, has no touchscreen, and lacks keyboard backlighting entirely (that feature lives one tier up in Apple’s lineup, on the MacBook Air). The Neo does have a headphone jack, which the XPS skips, and its trackpad registers clicks anywhere on its surface, while the XPS uses an older hinge-style design that only registers clicks near the bottom.
Does Intel still need to be plugged in to perform?
This is the most notable shift in the comparison. Intel laptops have historically needed wall power to hit their top performance tier, while Apple Silicon chips perform identically on battery or plugged in. In the Python multi-core Mandelbrot test, the XPS completed the benchmark in about 57.6 seconds plugged in and roughly 65 seconds on battery across repeated runs, both faster than the Neo’s 90-plus second times in either power state. That’s a real change from Intel’s previous XPS 13 generation, which ran a similar test in 80 seconds against a Qualcomm Snapdragon chip’s 122 seconds, meaning Intel improved its own numbers release over release while also closing the gap with Apple Silicon on unplugged performance.
How do they handle sustained, heavy workloads?
Short benchmarks don’t tell the whole story for developers running long compiles or training jobs. Running the multi-core test five times back to back revealed the tradeoff of each design. The fanless MacBook Neo slowed from 34.1 seconds to 36.6 seconds per run, about 7% degradation from thermal throttling, which is expected behavior for a chip protecting itself without active cooling. The XPS’s fan kept it from throttling in the same linear way, but its times bounced around more unpredictably (73, 71, 63, 67, 63 seconds) and the chassis ran warmer under load, with internal temperatures reported around 42°C compared to roughly 38°C on the Neo. The XPS stayed faster overall, but the inconsistency between runs is worth noting if predictable performance matters more than peak speed.
Which one is better for local LLM inference?
For small, practical local models, the two are close. With both tested through llama.cpp, the XPS slightly outprocessed the Neo on prompt reading speed (156 vs 142 tokens/sec) for a Qwen3 4B model at 4-bit quantization, while the Neo generated output tokens faster (17.9 vs 15.7 tokens/sec). Both numbers comfortably exceed typical human reading speed, so the practical difference is small for everyday chatbot-style use.
The real limitation on both machines is memory. With only 8GB of total system RAM, Llama CPP reported just 4.3GB available to the XPS’s integrated GPU and 5.3GB to the Neo’s, not enough headroom for an 8B parameter model even at 4-bit quantization, which already needs roughly 5GB just to load. When that model spilled into CPU and system RAM, the Neo still produced a usable 7.1 tokens per second, while the XPS dropped to 0.23 tokens per second, about one token every four seconds, likely due to disk swapping rather than CPU compute alone. For anyone planning to run local models seriously, this points to RAM capacity, not CPU architecture, as the real bottleneck, and neither 8GB configuration is enough for models beyond the 4B class.
Is the Dell XPS 13 or MacBook Neo the better buy for developers?
Price parity makes this closer than past comparisons. The XPS offers a sharper screen, a backlit keyboard, double the storage with the option to upgrade it, faster compile times whether plugged in or on battery, and a configuration path up to 16GB of RAM, something the Neo doesn’t offer. Apple’s advantage comes down to real-world efficiency: better battery life for equivalent work, no throttling surprises under short bursts, instant wake from sleep, and the integration benefits of being inside Apple’s ecosystem if you already use an iPhone or iPad.
Neither laptop is a clear loser. For someone deep in the Apple ecosystem, the Neo remains a straightforward choice. For developers who value compile speed, upgradeable storage, and a better screen at the same price, the XPS makes a genuinely strong case that wasn’t there with Intel’s previous generation.
Frequently Asked Questions
Is the MacBook Neo the same as a MacBook Air?
No. The MacBook Neo tested here uses Apple’s A18 Pro chip (the same silicon found in iPhone Pro models) rather than Apple’s M-series laptop chips, and it lacks a keyboard backlight, a feature reserved for the MacBook Air line.
Can either laptop run large local AI models well?
Not with 8GB of RAM. Both struggled once an 8B parameter model exceeded available GPU memory and spilled into system RAM, with the Neo staying usable at 7.1 tokens per second and the XPS dropping to a near-unusable 0.23 tokens per second. Models in the 4B parameter range at 4-bit quantization run fine on both.
Does the XPS 13 still need to be plugged in for best performance?
One coffee. One working app.
You bring the idea. Remy manages the project.
Not anymore, based on this testing. The XPS matched or beat its plugged-in benchmark times while running on battery, a reversal from older Intel laptops that depended on wall power for peak clock speeds.
Which laptop has better battery efficiency?
The MacBook Neo. In a combined Python compiling and LLM inference test, the Neo used about 13% of its battery while the XPS used about 21% to complete the same work, roughly 60% more battery drain for equal output.
Can you upgrade storage or RAM on either laptop?
The XPS 13’s SSD is removable and can be swapped for a larger drive, and it can be configured up to 16GB of RAM at purchase. The MacBook Neo’s 256GB storage is fixed permanently and its RAM cannot be upgraded.
