What is the biggest model the Apple M2 can run?

gpt-oss 20B, which needs about 16 GB, on the 24 GB version at about 18 tok/s.

What is the biggest model the Apple M4 can run?

Qwen3 32B, which needs about 22 GB, on the 32 GB version at about 3 tok/s.

Apple M2 vs Apple M4 for local AI

Q: Which is better for local AI, the Apple M2 or the Apple M4?

The Apple M4: it fits bigger models and runs faster. The Apple M4 fits more: up to Qwen3 32B, against gpt-oss 20B on the Apple M2. The Apple M4 generates about 20% faster on any model both can run, because it has 120 GB/s of memory bandwidth to the Apple M2's 100 GB/s.

The Apple M4 fits more: up to Qwen3 32B, against gpt-oss 20B on the Apple M2. The Apple M4 generates about 20% faster on any model both can run, because it has 120 GB/s of memory bandwidth to the Apple M2's 100 GB/s.

The numbers that matter

Memory and bandwidth

Memory decides which models fit. Bandwidth decides how fast they generate, because every token reads the model's active weights once.

	Apple M2	Apple M4
Memory	8 or 16 or 24 GB unified memory	16 or 24 or 32 GB unified memory
Usable for models	About 16 GB of 24 GB	About 21 GB of 32 GB
Bandwidth	100 GB/s	120 GB/s
Biggest model that fits	gpt-oss 20B	Qwen3 32B
Best everyday pick	Qwen3 4B, ~22 tok/s	Qwen3 30B-A3B, ~21 tok/s

Model by model

Every model, side by side

Estimated generation speed at the largest memory size each was sold in (24 GB and 32 GB).

Model	Apple M2	Apple M4
Qwen3 4B About 4 GB	~22 tok/s Usable	~26 tok/s Usable
Qwen3 8B About 7 GB	~11 tok/s Usable	~13 tok/s Usable
Gemma 3 12B About 10 GB	~7 tok/s Slow	~8 tok/s Slow
Qwen3 14B About 11 GB	~6 tok/s Slow	~7 tok/s Slow
gpt-oss 20B About 16 GB	~18 tok/s Usable	~21 tok/s Usable
Gemma 3 27B About 19 GB	Needs 19 GB	~4 tok/s Slow
Qwen3 30B-A3B About 21 GB	Needs 21 GB	~21 tok/s Usable
Qwen3 32B About 22 GB	Needs 22 GB	~3 tok/s Slow
Llama 3.3 70B About 45 GB	Needs 45 GB	Needs 45 GB
gpt-oss 120B About 67 GB	Needs 67 GB	Needs 67 GB
Qwen3 235B-A22B About 144 GB	Needs 144 GB	Needs 144 GB
DeepSeek R1 671B About 406 GB	Needs 406 GB	Needs 406 GB

Which should you get?

The Apple M4: it fits bigger models and runs faster. Price matters as much as either number: check used prices on the builds page, where every parts list has a measured or estimated speed.

How we estimate

Both columns use the same formula: memory bandwidth divided by the weights read per token, calibrated on our own rack, with Ollama's default 4-bit sizes. Real results are usually within about a third. Measured either one? Send us the number.

Compare

Other matchups

Every comparison