GPUs & Macs › Apple M4 (10-core GPU)
Apple M4 (10-core GPU): local LLM speed
Llama 2 7B (Q4_0) generates 24.1 tokens/s on the Apple M4 (10-core GPU) (Metal), and processes prompts at 221 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 24.1 | 221 | Q4_0 | Metal FA off | tg128 | llama.cpp #4167 details |
| Llama 2 7B | 13.5 | 224 | Q8_0 | Metal FA off | tg128 | llama.cpp #4167 details |
| Llama 2 7B | 7.4 | 230 | F16 | Metal FA off | tg128 | llama.cpp #4167 details |
| Llama 3.2 1B Instruct | 74.3 | 1,045 | Q4_K_M | llamafile | LocalScore avg · n=3 | LocalScore LocalScore LocalScore details |
| Llama 3.1 8B Instruct | 16.8 | 166 | Q4_K_M | llamafile | LocalScore avg · n=3 | LocalScore LocalScore LocalScore details |
| Qwen2.5 14B Instruct | 9.0 | 87 | Q4_K_M | llamafile | LocalScore avg · n=3 | LocalScore LocalScore LocalScore details |
What fits on the Apple M4 (10-core GPU)
Estimates for this chip are scaled from a measured run: Llama 2 7B F16 at 7.4 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 | 13.8 estimate |
| gemma-4-12B-it | Q6_K | 9.8 GB | 10.2 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 22.3 estimate |
| Qwen3.6-35B-A3B | UD-IQ1_M | 10.0 GB | 58.1 estimate |
| gemma-4-26B-A4B-it | UD-IQ2_M | 10.0 GB | 32.5 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 12.2 estimate |
With 36 GB unified memory
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | UD-Q6_K | 22.0 GB | 4.6 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 7.9 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 22.3 estimate |
| Qwen3.6-35B-A3B | UD-Q5_K_S | 24.9 GB | 23.4 estimate |
| gemma-4-26B-A4B-it | UD-Q6_K | 23.2 GB | 14.0 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 12.2 estimate |
With 64 GB unified memory
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | Q8_0 | 29.0 GB | 3.4 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 7.9 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 22.3 estimate |
| Qwen3.6-35B-A3B | Q8_0 | 36.9 GB | 15.8 estimate |
| gemma-4-26B-A4B-it | Q8_0 | 26.9 GB | 12.1 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 12.2 estimate |
With 128 GB unified memory
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | Q8_0 | 29.0 GB | 3.4 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 7.9 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 10.5 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 22.3 estimate |
| Qwen3.6-35B-A3B | Q8_0 | 36.9 GB | 15.8 estimate |
| Qwen3.8-Flash-Next | UD-IQ3_XXS | 82.0 GB | 1.2 estimate |
| gemma-4-26B-A4B-it | Q8_0 | 26.9 GB | 12.1 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 10.5 estimate |
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)