GPUs & Macs › Apple M5 Ultra (80-core GPU)

Apple M5 Ultra (80-core GPU): local LLM speed

Llama 2 7B (Q4_0) generates 179 tokens/s on the Apple M5 Ultra (80-core GPU) (Metal), and processes prompts at 4,945 tokens/s.

Measured results

Model	Generate tok/s	Prompt tok/s	Quant	Backend	Test	Source
Llama 2 7B	179	4,945	Q4_0	Metal FA off	tg128	llama.cpp #4167 details
Llama 2 7B	119	4,911	Q8_0	Metal FA off	tg128	llama.cpp #4167 details
Llama 2 7B	73.6	5,207	F16	Metal FA off	tg128	llama.cpp #4167 details

Vendor	Apple
Memory	Unified (depends on configuration)
Memory bandwidth	1228.8 GB/s
GPU cores	80
Spec source	Wikipedia (Apple M5)
Spec note	DERIVED: 2x M5 Max (Wikipedia lists LPDDR5X-9600; no explicit Ultra bandwidth cell); matches llama.cpp #4167 table (1228)

Check prices

Amazon

As an Amazon Associate I earn from qualifying purchases.

Rent one by the hour

Rent a GPU on RunPod

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.

What fits on the Apple M5 Ultra (80-core GPU)

Estimates for this chip are scaled from a measured run: Llama 2 7B F16 at 73.6 tok/s (Metal, source). Fit assumes the whole model in GPU memory with ~8K context.

With 16 GB unified memory

Model	Best quant that fits	File	Expected tok/s
Qwen3.8-27B	UD-IQ2_XXS	7.3 GB	136 estimate
gemma-4-12B-it	Q6_K	9.8 GB	101 estimate
Qwen3.5-9B	Q8_0	9.5 GB	104 estimate
Qwen3.5-4B	Q8_0	4.5 GB	221 estimate
Qwen3.6-35B-A3B	UD-IQ1_M	10.0 GB	576 estimate
gemma-4-26B-A4B-it	UD-IQ2_M	10.0 GB	322 estimate
MiMo-V2.6-Distill-Qwen-9B	Q8_0	9.5 GB	104 estimate
gemma-4-E4B-it	Q8_0	8.2 GB	121 estimate

With 36 GB unified memory

Model	Best quant that fits	File	Expected tok/s
Qwen3.8-27B	UD-Q6_K	22.0 GB	45.1 estimate
gemma-4-12B-it	Q8_0	12.7 GB	78.3 estimate
Qwen3.5-9B	Q8_0	9.5 GB	104 estimate
Qwen3.5-4B	Q8_0	4.5 GB	221 estimate
Qwen3.6-35B-A3B	UD-Q5_K_S	24.9 GB	232 estimate
gemma-4-26B-A4B-it	UD-Q6_K	23.2 GB	139 estimate
MiMo-V2.6-Distill-Qwen-9B	Q8_0	9.5 GB	104 estimate
gemma-4-E4B-it	Q8_0	8.2 GB	121 estimate

With 64 GB unified memory

Model	Best quant that fits	File	Expected tok/s
Qwen3.8-27B	Q8_0	29.0 GB	34.1 estimate
gemma-4-12B-it	Q8_0	12.7 GB	78.3 estimate
Qwen3.5-9B	Q8_0	9.5 GB	104 estimate
Qwen3.5-4B	Q8_0	4.5 GB	221 estimate
Qwen3.6-35B-A3B	Q8_0	36.9 GB	157 estimate
gemma-4-26B-A4B-it	Q8_0	26.9 GB	120 estimate
MiMo-V2.6-Distill-Qwen-9B	Q8_0	9.5 GB	104 estimate
gemma-4-E4B-it	Q8_0	8.2 GB	121 estimate

With 128 GB unified memory

Model	Best quant that fits	File	Expected tok/s
Qwen3.8-27B	Q8_0	29.0 GB	34.1 estimate
gemma-4-12B-it	Q8_0	12.7 GB	78.3 estimate
Qwen3.5-9B	Q8_0	9.5 GB	104 estimate
Qwen3.5-4B	Q8_0	4.5 GB	221 estimate
Qwen3.6-35B-A3B	Q8_0	36.9 GB	157 estimate
Qwen3.8-Flash-Next	UD-IQ3_XXS	82.0 GB	12.1 estimate
gemma-4-26B-A4B-it	Q8_0	26.9 GB	120 estimate
MiMo-V2.6-Distill-Qwen-9B	Q8_0	9.5 GB	104 estimate

Open in the picker

Similar hardware

Apple M1 (7-core GPU) · Apple M1 (8-core GPU) · Apple M1 Pro (14-core GPU) · Apple M1 Pro (16-core GPU) · Apple M1 Max (24-core GPU) · Apple M1 Max (32-core GPU) · Apple M1 Ultra (48-core GPU) · Apple M1 Ultra (64-core GPU)