Apple M4 Pro
LaptopApple unified-memory SoC: up to 64 GB of unified memory at 273 GB/s, roughly 75% more bandwidth than M3 Pro. Used in MacBook Pro and Mac mini. Apple does not publish a TDP, so that column is left empty.
41
Run records
18
Models covered
4
Frameworks covered
L0 Self-reported × 7 · L2 Cross-framework verified × 34
Evidence mix
Typical measured performance
Published benchmark averages per model × framework × quantization on this hardware.
- Ollama · Q4_K_S Decode 136.48 · Prefill 705.2 ·
- Strata · IQ2_M Decode 177.44 · Prefill 840.26 ·
- llama.cpp · IQ4_XS Decode 286.93 · Prefill 1431.45 ·
- llama.cpp · Q5_K_M Decode 205.09 · Prefill 1101.85 ·
- Strata · Q5_K_S Decode 123.85 · Prefill 643.4 ·
- vLLM · Q3_K_M Decode 131.41 · Prefill 579.15 ·
- Ollama · Q8_0 Decode 61.9 · Prefill 265.27 ·
- vLLM · Q5_K_M Decode 115.86 · Prefill 428.93 ·
- Ollama · IQ2_XS Decode 164.02 · Prefill 802.68 ·
- Ollama · Q8_0 Decode 86.82 · Prefill 468.6 ·
- Strata · Q2_K Decode 389.55 · Prefill 1966.41 ·
- llama.cpp · IQ4_XS Decode 138.37 · Prefill 661.86 ·
- llama.cpp · Q4_K_S Decode 123.54 · Prefill 441.3 ·
- Strata · IQ3_XS Decode 218.39 · Prefill 1145.06 ·
- Strata · Q3_K_L Decode 206.28 · Prefill 783.16 ·
- Strata · IQ1_S Decode 325.5 · Prefill 1591.11 ·
- Strata · IQ2_M Decode 303.45 · Prefill 1578.72 ·
- vLLM · IQ4_XS Decode 172.49 · Prefill 661.58 ·
- vLLM · Q3_K_S Decode 184.21 · Prefill 1008.8 ·
- Ollama · Q4_K_S Decode 154.32 · Prefill 813.63 ·
- Ollama · Q5_K_M Decode 114.09 · Prefill 438.42 ·
- vLLM · Q5_K_M Decode 270.98 · Prefill 944.94 ·
- vLLM · IQ2_M Decode 64.88 · Prefill 330.09 ·
- vLLM · IQ3_XS Decode 87.63 · Prefill 428.03 ·
- vLLM · Q4_K_S Decode 72.54 · Prefill 364.75 ·
- Ollama · Q5_K_M Decode 56.24 · Prefill 299.52 ·
- Ollama · Q5_K_S Decode 59.26 · Prefill 228.14 ·
- llama.cpp · IQ2_M Decode 103.7 · Prefill 452.45 ·
- llama.cpp · Q5_K_S Decode 68.46 · Prefill 358.45 ·
- vLLM · Q2_K Decode 110.03 · Prefill 407.67 ·
- vLLM · Q4_K_S Decode 90.85 · Prefill 484.54 ·
- vLLM · Q3_K_S Decode 87.73 · Prefill 414.18 ·
- vLLM · Q4_K_M Decode 76.77 · Prefill 324.87 ·
- Strata · IQ3_S Decode 124.6 · Prefill 542.9 ·
- llama.cpp · IQ2_XS Decode 0.8 · Prefill 6 ·
- llama.cpp · IQ3_S Decode 0.7 · Prefill 5.3 ·
- llama.cpp · IQ2_XS Decode 0.8 · Prefill 6 ·
- llama.cpp · IQ3_S Decode 0.7 · Prefill 5.3 ·
- llama.cpp · IQ2_XS Decode 23.4 · Prefill 175.5 ·
- llama.cpp · Q5_K_M Decode 10.6 · Prefill 79.5 ·
- llama.cpp · Q4_K_M Decode 13 · Prefill 97.5 ·
Fit check
Physics-based estimate from VRAM and memory bandwidth: pick a model to see which quantizations fit and the expected speed
Formula: tok/s = (bandwidth ÷ model size) × 0.55, adapted from llmfit; see run records for measured numbers
Key specs
- Vendor
- Apple
- Architecture
- M4 Pro
- VRAM
- 64 GB
- Memory bandwidth
- 273 GB/s
- Release date
- 2024-10-30
Run records
41 configurations
| Model | Framework | Quantization | Decode | Gen time | Samples |
|---|---|---|---|---|---|
| Qwen3.5-4B | Ollama | Q4_K_S | 136.5 tok/s | — | 1 |
| Qwen3.5-9B | Strata | IQ2_M | 177.4 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | IQ4_XS | 286.9 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | Q5_K_M | 205.1 tok/s | — | 1 |
| Qwen3.8-9B | Strata | Q5_K_S | 123.8 tok/s | — | 1 |
| Qwen3-8B | vLLM | Q3_K_M | 131.4 tok/s | — | 1 |
| Qwen3-8B | Ollama | Q8_0 | 61.9 tok/s | — | 1 |
| Qwen3-4B | vLLM | Q5_K_M | 115.9 tok/s | — | 1 |
| Qwen3-4B | Ollama | IQ2_XS | 164 tok/s | — | 1 |
| Qwen3-4B | Ollama | Q8_0 | 86.8 tok/s | — | 1 |
| Qwen3-1.7B | Strata | Q2_K | 389.6 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | llama.cpp | IQ4_XS | 138.4 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | llama.cpp | Q4_K_S | 123.5 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | Strata | IQ3_XS | 218.4 tok/s | — | 1 |
| Qwen3-4B-Thinking-2507 | Strata | Q3_K_L | 206.3 tok/s | — | 1 |
| Qwen2.5-0.5B-Instruct | Strata | IQ1_S | 325.5 tok/s | — | 1 |
| Qwen2.5-0.5B-Instruct | Strata | IQ2_M | 303.4 tok/s | — | 1 |
| Qwen2.5-Omni-3B | vLLM | IQ4_XS | 172.5 tok/s | — | 1 |
| Qwen2.5-Omni-3B | vLLM | Q3_K_S | 184.2 tok/s | — | 1 |
| Qwen2.5-Omni-3B | Ollama | Q4_K_S | 154.3 tok/s | — | 1 |
| Qwen2.5-Omni-3B | Ollama | Q5_K_M | 114.1 tok/s | — | 1 |
| Qwen3-Embedding-0.6B | vLLM | Q5_K_M | 271 tok/s | — | 1 |
| gemma-4-31B-it | vLLM | IQ2_M | 64.9 tok/s | — | 1 |
| gemma-4-26B-A4B-it | vLLM | IQ3_XS | 87.6 tok/s | — | 1 |
| gemma-4-26B-A4B-it | vLLM | Q4_K_S | 72.5 tok/s | — | 1 |
| gemma-4-26B-A4B-it | Ollama | Q5_K_M | 56.2 tok/s | — | 1 |
| gemma-4-26B-A4B-it | Ollama | Q5_K_S | 59.3 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | IQ2_M | 103.7 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | Q5_K_S | 68.5 tok/s | — | 1 |
| gemma-4-12B-it | vLLM | Q2_K | 110 tok/s | — | 1 |
| gemma-4-12B-it | vLLM | Q4_K_S | 90.8 tok/s | — | 1 |
| DeepSeek-R1-Distill-Qwen-14B | vLLM | Q3_K_S | 87.7 tok/s | — | 1 |
| DeepSeek-R1-Distill-Qwen-14B | vLLM | Q4_K_M | 76.8 tok/s | — | 1 |
| DeepSeek-R1-Distill-Qwen-14B | Strata | IQ3_S | 124.6 tok/s | — | 1 |
| GLM-5.3 | llama.cpp | IQ2_XS | 0.8 tok/s | — | 1 |
| GLM-5.3 | llama.cpp | IQ3_S | 0.7 tok/s | — | 1 |
| DeepSeek-V4.1-Flash | llama.cpp | IQ2_XS | 0.8 tok/s | — | 1 |
| DeepSeek-V4.1-Flash | llama.cpp | IQ3_S | 0.7 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | IQ2_XS | 23.4 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | Q5_K_M | 10.6 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | Q4_K_M | 13 tok/s | — | 1 |