The Specifications miss the Memory.
I'm not an APPLE user (yet?), but no fan is nice. Tho it looks like the performance throttles to 1/3 (-66%) when stress-tested (see the graph above the "Speakers" section).
3.33 lbs/1.512 kg for 15.30 inch is just ok and nothing special.
The pixel response times are shower than vs predecessor Air 15 M4, but are still (just) ok and nice color coverage and brightness.
29% faster in Cyberpunk 2077 vs predecessor is good, but this is still only 23.3 FPS and overall the iGPU is 1/3 to 1/2 as fast as it should be at the minimum. Wonder if this was tested on a cool system, as when stress-tested, the performance drops to 1/3.
Wish the iGPU was faster/better utilized/more efficient:
Steel Nomad (Vulkan):
- Radeon 890M: Average score: 753
- Bandwidth: 89.6 GB/s (= 128-bit * 5600 MT/s / 1000 / 8)
- -> FPS score per GB/s = 753/89.6 = 8.40
- Apple MacBook Air 15 M5: 1070
- Bandwidth: 153.6 GB/s (= 128-bit * 9600 MT/s / 1000 / 8)
- -> FPS score per GB/s = 6.97
- 4050 Laptop: Average score: 1789
- Bandwidth: 192.0 GB/s (= 96-bit * 16 Gbit/s / 8)
- -> FPS score per GB/s = 9.32
This doesn't take into account the size of L2 cache and system level caches, which is one of the reasons one architecture maybe able to get away with less DDR bandwidth over another but still get as high or higher FPS score.
There's just too many other variables which impact gfx performance to make generalisations like this.
But in general, efficiency goes up with more cores. This is apples smallest GPU. What do you think is going to happen once the bigger Pro and Max variants that have been recently released are tested?
Is there a reason you insist on doing calculations based on synthetic benchmarks when we have in-game tests and scores?
It's throttling under combined synthetic CPU+GPU stress test with no fan. It doesn't throttle when gaming.
> I'm not an APPLE user (yet?), but no fan is nice.
it is not nice, it is a game changer. probably I have Asperger's, but it reduces my stress levels like nothing else. It just brings zen.
> The CPU performance is not reduced on battery power.
This is not emphasized enough. The performance is out of this world at the given TDP. But that's not all, Windows laptops are usually benchmarked with high performance power profiles. But virtually no one uses them with such profiles, especially not on battery power - since the runtimes would go to complete s*** and it would be louder than a plane.
So not only do you have the best performance on paper for most tasks at amazing efficiency and power draw, but the practical experience of using the laptop will be much better to what the numbers suggest - since again, Intel/AMD (Qualcomm to a lesser extent but still significantly) throttle like crazy on battery power.
This is not emphasized enough honestly. There is no competition and the gap in the actual user experience, especially on battery power, is MUCH bigger than what the benchmarks suggest.
Indirectly, this also means that battery life benchmarks are not fair.
You will tear your hairs out while using heavily throttled Intel CPUs on Windows laptops, apps will be slow to open, tabs will be slow to switch etc. Unless you switch to high perf power profiles, which will reduce battery life by 80%.
But of course, on Macbooks, there is no such benchmark hacking. The experience is snappy throughout. And even with that, they can still deliver 17 hours of battery life. Nothing with a comparable user experience comes close in the non-apple laptop market.
Fanless is not really an attraction for me. What I want is a quiet design. I'm not bothered by a fan that I can't hear. Windows OEMs are incredibly incompetent and can't even achieve giving that as an option.
Quote from: JimD on March 09, 2026, 12:11:02Fanless is not really an attraction for me.
What a reliability? Would you consider that an important factor to you? Moving parts are considered less reliable. That's one of the reasons why we moved from HDDs to SSDs (apart from massive speed boost in read seek and write times) and also moved away from optical mediums (CDs, DVDs, BluRays) in laptops.
Fanless devices tend to have longer runtimes as well.
Quote from: dumb_oems on March 09, 2026, 11:44:29This is not emphasized enough
...
the actual user experience, especially on battery power
I agree. I don't think people understand. There are many gaming laptops out there that give wonderful performance @ almost 60 dB noise. So basically like an airplane taking of a runway. Is that really usable? Then there's the heat output, enough to make someone impotent by sterilizing the sperm on your ballsac.
What happens to your thermals / noise output if you live in a warmer country which is especially becoming increasingly prevalent in this era of global warming?
We're not even getting into battery health and longevity increasing due to less heat.
My hope is that this showcase is enough of an embarrassment that the other big tech entities (Nvidia, AMD, Intel, Qcomm, etc) in the windows ecosystem are pretty much forced to wake up, start innovating again and eventually catch up in few years so we can all be benefit. Much like how over a decade ago Apple were among the first to start mandating SSDs & IPS displays in laptops when HDDs and TN were the norm.
Can you please check the screen for Dithering? thank you.
screen contrast is lower so much than m4 15 macbook air ,is a defective one or m5 just had bad panel, we need more example
Will the 24 GB RAM option fit Qwen3.5-27B-UD-Q4_K_XL.gguf (17.6 GB) or Qwen3.5-27B-UD-Q5_K_XL.gguf (20.2 GB)? (huggingface.co/unsloth/Qwen3.5-27B-GGUF) (I know there's mlx-community/Qwen3.5-27B-4bit (16.1 GB) too, but I don't know if its perplexity is good)
Quote from: Will MBAir fit 27B quants on March 11, 2026, 12:23:33Will the 24 GB RAM option fit Qwen3.5-27B-UD-Q4_K_XL.gguf (17.6 GB) or Qwen3.5-27B-UD-Q5_K_XL.gguf (20.2 GB)? (huggingface.co/unsloth/Qwen3.5-27B-GGUF) (I know there's mlx-community/Qwen3.5-27B-4bit (16.1 GB) too, but I don't know if its perplexity is good)
I have a 15" M2 Air with 24GB RAM, but it can only run 3B models max because of overheating. Work is fine, gaming is fine, but LLMs throttle it to 0.4GHz on GPU in a minute. Sorry, you would need something with a fan or two to make those models useful.
Quote from: not_anton on March 11, 2026, 17:16:44Quote from: Will MBAir fit 27B quants on March 11, 2026, 12:23:33Will the 24 GB RAM option fit Qwen3.5-27B-UD-Q4_K_XL.gguf (17.6 GB) or Qwen3.5-27B-UD-Q5_K_XL.gguf (20.2 GB)? (huggingface.co/unsloth/Qwen3.5-27B-GGUF) (I know there's mlx-community/Qwen3.5-27B-4bit (16.1 GB) too, but I don't know if its perplexity is good)
I have a 15" M2 Air with 24GB RAM, but it can only run 3B models max because of overheating. Work is fine, gaming is fine, but LLMs throttle it to 0.4GHz on GPU in a minute. Sorry, you would need something with a fan or two to make those models useful.
Good to know, but will those quants fit (if I had ot guess I'd say yes)? There's also sysctl iogpu.wired_limit_mb=<MB> (I know not to assign too much to the VRAM (1-3 GB may be ok), as the OS may start to write/swap to the SSD).
12W average in idle?? how can that be, lunar lake is only a third of this.
why are you not doing the battery test with idle and video playback any more??
and still no matte option for the screen, so idiotic without any reason.
until then i wont consider an air.
Quote from: juri on March 13, 2026, 03:09:11and still no matte option for the screen, so idiotic without any reason.
It's not 12W, look again. Maybe the matte coating on your screen played a prank on you ;)
And I'm so glad that it's glossy (richer colors, sharper text). Mind that while it's glossy, the anti-reflective coating/ability is very good.
This may be interesting for reviewers:
Quote from: youtube.com/watch?v=HKxIGgyeISMApple's Energy Model - Deconstructed
In this video, I reverse engineer Apple's Energy Model on the Mac Studio M4 Max. In the process I explain why and how measured DC power can appear up to 3 times higher than reported M4 Max GPU power.
[..]
As expected, a vid by Alex Ziskind (youtube.com/watch?v=XGe7ldwFLSE), in this case, proves that Apples's claim of
Quote from: en.wikipedia.org/wiki/Apple_M5#PerformancePeak GPU AI compute: over 4× faster
does not apply to running 3rd party LLMs and only the RAM/unified memory bandwidth increase increases the token generation (28% = 1.28 = 153.6 GB/s (M5) / 120 GB/s (M4)). (153.6 GB/s = 128-bit * 9600 MT/s / 1000 / 8)
Or look up the 153.6 GB/s and 120 GB/s values here: en.wikipedia.org/wiki/Apple_silicon#M-series_SoCs.
First I confirmed that the CPU and GPU scores of my new Air 15 M5 align with what is expected.
Here are some llama.cpp's llama-bench benchmarks:
First with battery saving mode on:
/Users/../llama-b8740/llama-bench --no-warmup -m /Users/../Qwen3.5-9B-UD-Q4_K_XL.gguf -p 128 -n 256 -t 1,2,3,4
ggml_metal_device_init: testing tensor API for f16 support
ggml_metal_library_init_from_source: error compiling source
ggml_metal_device_init: - the tensor API is not supported in this environment - disabling
ggml_metal_library_init: using embedded metal library
ggml_metal_library_init: loaded in 0.013 sec
ggml_metal_rsets_init: creating a residency set collection (keep_alive = 180 s)
ggml_metal_device_init: GPU name: MTL0
ggml_metal_device_init: GPU family: MTLGPUFamilyApple10 (1010)
ggml_metal_device_init: GPU family: MTLGPUFamilyCommon3 (3003)
ggml_metal_device_init: GPU family: MTLGPUFamilyMetal4 (5002)
ggml_metal_device_init: simdgroup reduction = true
ggml_metal_device_init: simdgroup matrix mul. = true
ggml_metal_device_init: has unified memory = true
ggml_metal_device_init: has bfloat = true
ggml_metal_device_init: has tensor = false
ggml_metal_device_init: use residency sets = true
ggml_metal_device_init: use shared buffers = true
ggml_metal_device_init: recommendedMaxWorkingSetSize = 12713.12 MB
| model | size | params | backend | threads | test | t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | pp128 | 110.11 ± 4.18 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | tg256 | 9.50 ± 0.30 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | pp128 | 107.58 ± 6.53 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | tg256 | 9.45 ± 0.08 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | pp128 | 110.79 ± 1.14 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | tg256 | 9.29 ± 0.09 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | pp128 | 110.78 ± 1.42 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | tg256 | 8.75 ± 0.95 |With the default no battery saving mode:
/Users/../llama-b8740/llama-bench --no-warmup -m /Users/../Qwen3.5-9B-UD-Q4_K_XL.gguf -p 128 -n 256 -t 1,2,3,4
[..]
| model | size | params | backend | threads | test | t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | pp128 | 216.04 ± 8.47 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | tg256 | 20.17 ± 0.04 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | pp128 | 209.43 ± 0.35 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | tg256 | 20.21 ± 0.03 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | pp128 | 196.55 ± 5.79 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | tg256 | 18.34 ± 3.22 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | pp128 | 193.50 ± 2.13 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | tg256 | 19.57 ± 0.18 |
On battery saving mode (for no battery saving mode, you guessed it, just like above, multiply by ~2):
/Users/../llama-b8740/llama-bench --no-warmup -m /Users/../Qwen3.5-9B-UD-Q4_K_XL.gguf -p 64 -n 64 -t 1-12
| model | size | params | backend | threads | test | t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | pp64 | 107.20 ± 6.39 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 1 | tg64 | 9.75 ± 0.06 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | pp64 | 109.29 ± 0.92 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 2 | tg64 | 9.62 ± 0.01 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | pp64 | 109.05 ± 0.96 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 3 | tg64 | 9.47 ± 0.00 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | pp64 | 108.81 ± 0.81 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 4 | tg64 | 9.23 ± 0.13 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 5 | pp64 | 108.59 ± 0.62 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 5 | tg64 | 9.20 ± 0.01 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 6 | pp64 | 108.57 ± 1.21 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 6 | tg64 | 9.12 ± 0.01 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 7 | pp64 | 108.34 ± 1.24 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 7 | tg64 | 9.02 ± 0.04 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 8 | pp64 | 108.07 ± 1.57 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 8 | tg64 | 8.97 ± 0.01 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 9 | pp64 | 107.81 ± 0.76 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 9 | tg64 | 8.76 ± 0.15 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 10 | pp64 | 107.75 ± 0.58 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 10 | tg64 | 8.82 ± 0.03 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 11 | pp64 | 107.64 ± 0.44 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 11 | tg64 | 8.71 ± 0.03 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 12 | pp64 | 107.18 ± 0.34 |
| qwen35 9B Q4_K - Medium | 5.55 GiB | 8.95 B | MTL,BLAS | 12 | tg64 | 8.67 ± 0.05 |
This confirms that the number of threads doesn't matter. On my desktop 7800X3D 4 threads gives pretty much the fastest tokens per second.
You got the screen resolution wrong. It is 2880 x 1864.