Can an Apple M4 Pro 48GB run Qwen2.5 1.5B Instruct?

Yes

Yes. At Q8_0 it needs 2.14 GB of the 36 GB available and generates around 122.2 tokens per second at an 8k context window.

The numbers

Qwen2.5 1.5B Instruct has 1.54 billion parameters across 28 layers. The Apple M4 Pro 48GB has 48 GB of unified memory at 273 GB/s, of which about 36 GB is left for a model once the operating system has taken its share.

QuantisationTotal neededFitsSpeed
Q8_0 2.14 GB yes 122.2 t/s
Q6_K 1.79 GB yes 152.7 t/s
Q5_K_M 1.63 GB yes 171.9 t/s
Q4_K_M 1.48 GB yes 196.3 t/s
IQ4_XS 1.38 GB yes 217.1 t/s
Q3_K_M 1.31 GB yes 231.5 t/s
IQ3_XXS degraded 1.16 GB yes 277.5 t/s
Q2_K degraded 1.21 GB yes 259.9 t/s
IQ2_XXS degraded 0.98 GB yes 362.1 t/s

9 of the 9 levels fit. Levels marked degraded are listed for completeness, not as advice.

How long a conversation it holds

At Q8_0, memory rises with the length of the conversation because the attention cache keeps a key and value for every token. The longest window that still fits on this device is 32k tokens.

ContextCacheTotalFits
4k 0.11 GB 1.98 GB yes
8k 0.22 GB 2.14 GB yes
16k 0.44 GB 2.45 GB yes
32k 0.88 GB 3.08 GB yes

See how fast it feels

Apple M4 Pro 48GB running Qwen2.5 1.5B Instruct at Q8_0

Wait for the first word571 ms
Then writes at122.2 tok/s
Whole answer2.0 s

YouWhy does my model use more memory when the conversation gets longer?

Model

Because of the KV cache. Every token you send leaves behind a key and a value vector in each layer of the model, and those stay in memory for as long as the conversation lasts.

The weights are a fixed cost: load a 4-bit 8B model and that is about 4.8 GB, whether you write one word or ten thousand. The cache is the part that grows, and it grows in a straight line with the number of tokens in the window.

How steeply depends on the model's attention design. With grouped-query attention, several query heads share one key-value pair, which cuts the cache by that ratio. Without it, every head keeps its own, and a long context can cost more memory than the weights themselves.

If you are short on memory, the first thing to try is lowering the context window in your runtime, not the quantisation.

Simulated from our estimate at a 512-token question, not a recording. Assumes nothing else is competing for the GPU. The same answer on RTX 4090 at Q8_0 would take about 377 ms.

Keep going

Everything the Apple M4 Pro 48GB runs, the full breakdown for Qwen2.5 1.5B Instruct, or check a different pairing in the calculator.