llamaperf

A18 Pro 8GB

APPLE · 8GB unified memory · 1 report

Is this your card?

On the A18 Pro 8GB, LFM3 1.7B at 4-bit, the model most run on it, generates a median of 40.0 tokens per second from one plain run (community reports, as of 22 Nov 2026).

As of 22 Nov 2026, the models most run on the A18 Pro 8GB, each one size at one quant level, with the median generation speed of its plain runs (one device, one request, no speculative decoding):

Bars are medians on one scale, 0 to 40.0 tokens/s. Press ▶ to watch text arrive at a row's speed.

Engines people use on it: MLX 1 (reports)

Run models on your A18 Pro 8GB? Add your numbers.

Your coding agent can run a speed test on your machine and send it here with a free account. Or paste a result you've already posted.

Reported performance on the A18 Pro 8GB

These reports mix GPU counts and memory setups, so a model listed here may have needed extra cards or CPU offloading. The calculator estimates which models fit in 8 GB of unified memory. Prompt processing is how fast the input is read, not how long you wait for the first word. How to measure time to first token.

One report on this card so far, so treat the figures as early. Add yours.

Speeds are as each source reported them. GPU count, offloading and concurrent requests change them, and prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing two reports.

LFM3 1.7B

A18 Pro 8GB · MLX · 8,192 ctx

Tone: positiveApple hardware
generation:
40.0 tokens/s
quant:
4-bit (MLX)
vision

User reports LFM3-1.7B at about 40 tok/s on an iPhone 16 Pro, with first token under 0.5 s. Setup is MLX 4-bit on the Apple GPU; the export is 1.07 GB and context goes to 8192 tokens. LFM3-VL-2B with a photo takes about 9 s to first token and about 20 tok/s decode, with a text follow-up starting in about 5 s. A CPU build (XNNPACK 8da4w, 1.17 GB) runs on iOS and Android.

Oct 30, 2026

Get a weekly email of new A18 Pro 8GB reports and newly released models that fit it.

Email me new reports

Reported configurations

Individual observations from this page, not expected speeds or a ranking. Select a model to open its report, with the source and offloading conditions. Reported context may be a limit rather than actual input length.

Model / hardwareQuant / engineReported contextGeneration
LFM3 1.7B
A18 Pro 8GB
4-bit
MLX
8,19240.0 tokens/s

Frequently asked

How fast is the A18 Pro 8GB for local LLMs?

On the A18 Pro 8GB, LFM3 1.7B at 4-bit, the model most run on it, generates a median of 40.0 tokens per second from one plain run (community reports, as of 22 Nov 2026).