GPUs & Macs › Apple A18 Pro (5-core GPU)
Apple A18 Pro (5-core GPU): local LLM speed
2 measured results for the Apple A18 Pro (5-core GPU).
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 3.2 1B Instruct | 33.5 | 429 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Llama 3.1 8B Instruct | 15.8 | 116 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the Apple A18 Pro (5-core GPU)
Estimates for this chip are scaled from a measured run: Llama 3.1 8B Instruct Q4_K_M at 15.8 tok/s (llamafile, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| gemma-4-12B-it | IQ4_XS | 6.4 GB | 12.2 estimate |
| Qwen3.5-9B | Q5_K_S | 6.4 GB | 12.2 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 17.4 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q4_K_M | 5.8 GB | 13.3 estimate |
| gemma-4-E4B-it | Q5_K_M | 5.5 GB | 14.2 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 18.2 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 64.9 estimate |
| gemma-4-E2B-it | Q8_0 | 5.0 GB | 15.7 estimate |
| DeepSeek-R1-0528-Qwen3-8B | Q5_K_M | 5.9 GB | 13.3 estimate |
| Llama-3.2-3B-Instruct | Q8_0 | 3.4 GB | 22.8 estimate |
| Phi-4-mini-instruct | Q8_0 | 4.1 GB | 19.1 estimate |
| gemma-3-4b-it | Q8_0 | 4.1 GB | 18.9 estimate |
| Qwen3-14B | UD-IQ3_XXS | 6.0 GB | 13.0 estimate |
| gemma-3-12b-it | Q3_K_M | 6.0 GB | 13.0 estimate |