Apple M4 Max
LaptopApple unified-memory SoC: up to 128 GB of unified memory at 546 GB/s (410 GB/s on the binned 32-core-GPU variant). Used in MacBook Pro and Mac Studio. Apple does not publish a TDP, so that column is left empty.
50
Run records
14
Models covered
4
Frameworks covered
L0 Self-reported × 3 · L2 Cross-framework verified × 47
Evidence mix
Typical measured performance
Published benchmark averages per model × framework × quantization on this hardware.
- vLLM · Q5_K_S Decode 122.2 · Prefill 481.39 ·
- Strata · IQ2_M Decode 265.43 · Prefill 1342.3 ·
- Strata · IQ3_S Decode 228.14 · Prefill 1118.4 ·
- llama.cpp · IQ2_M Decode 189.88 · Prefill 951.61 ·
- llama.cpp · IQ2_XS Decode 191.16 · Prefill 1032.54 ·
- vLLM · Q3_K_M Decode 174.24 · Prefill 673.46 ·
- Ollama · Q6_K Decode 99.52 · Prefill 397.09 ·
- vLLM · IQ4_XS Decode 115.28 · Prefill 541.34 ·
- vLLM · Q4_K_S Decode 115.9 · Prefill 629 ·
- Ollama · IQ2_XS Decode 138.01 · Prefill 631.53 ·
- llama.cpp · IQ3_XS Decode 121.63 · Prefill 540.05 ·
- llama.cpp · Q4_K_S Decode 111.01 · Prefill 387.61 ·
- Ollama · Q3_K_L Decode 102.3 · Prefill 395.71 ·
- llama.cpp · Q5_K_M Decode 206.25 · Prefill 667.88 ·
- Ollama · Q3_K_L Decode 222.79 · Prefill 746.94 ·
- Ollama · Q4_K_M Decode 207.82 · Prefill 1018.29 ·
- Strata · IQ4_XS Decode 330.49 · Prefill 1150.53 ·
- Strata · Q2_K Decode 349.71 · Prefill 1794.35 ·
- llama.cpp · IQ3_S Decode 165.6 · Prefill 722.22 ·
- vLLM · Q3_K_M Decode 166.04 · Prefill 735.49 ·
- Ollama · Q3_K_S Decode 165.47 · Prefill 649.98 ·
- Ollama · Q5_K_S Decode 111.59 · Prefill 447.41 ·
- llama.cpp · IQ4_XS Decode 313.09 · Prefill 1177.06 ·
- llama.cpp · Q4_K_S Decode 304.54 · Prefill 1527.8 ·
- Strata · Q2_K Decode 352.85 · Prefill 1792.79 ·
- Strata · Q3_K_L Decode 303.63 · Prefill 1479.8 ·
- vLLM · IQ2_XS Decode 251.74 · Prefill 1248.2 ·
- Ollama · Q8_0 Decode 103.18 · Prefill 552.91 ·
- Strata · IQ2_XS Decode 353.9 · Prefill 1746.97 ·
- Strata · IQ3_S Decode 255.89 · Prefill 1167.45 ·
- llama.cpp · Q2_K Decode 238 · Prefill 1037.55 ·
- llama.cpp · Q3_K_L Decode 189.78 · Prefill 777.5 ·
- vLLM · FP16 Decode 103.62 · Prefill 346.91 ·
- vLLM · IQ1_S Decode 287.05 · Prefill 1522.92 ·
- Ollama · FP16 Decode 81 · Prefill 424.01 ·
- Ollama · IQ4_XS Decode 182.06 · Prefill 996.81 ·
- vLLM · IQ3_S Decode 318.24 · Prefill 1490.26 ·
- vLLM · Q5_K_S Decode 322.68 · Prefill 1251.16 ·
- Strata · IQ2_XS Decode 304.05 · Prefill 1416.13 ·
- llama.cpp · Q3_K_M Decode 93.03 · Prefill 389.72 ·
- llama.cpp · Q8_0 Decode 53.15 · Prefill 271.21 ·
- vLLM · Q2_K Decode 111.52 · Prefill 400.99 ·
- vLLM · Q8_0 Decode 54.33 · Prefill 226.48 ·
- llama.cpp · IQ2_M Decode 187.13 · Prefill 653.55 ·
- llama.cpp · Q4_K_M Decode 140.93 · Prefill 615.86 ·
- vLLM · IQ3_XS Decode 181.71 · Prefill 766.49 ·
- Strata · Q2_K Decode 270.81 · Prefill 1355.35 ·
- llama.cpp · IQ3_S Decode 3.7 · Prefill 27.8 ·
- llama.cpp · IQ3_S Decode 5.8 · Prefill 43.5 ·
- llama.cpp · Q4_K_M Decode 26 · Prefill 195 ·
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 Max
- VRAM
- 128 GB
- Memory bandwidth
- 546 GB/s
- Release date
- 2024-10-30
Run records
50 configurations
| Model | Framework | Quantization | Decode | Gen time | Samples |
|---|---|---|---|---|---|
| Qwen3.8-27B | vLLM | Q5_K_S | 122.2 tok/s | — | 1 |
| Qwen3.8-27B | Strata | IQ2_M | 265.4 tok/s | — | 1 |
| Qwen3.8-27B | Strata | IQ3_S | 228.1 tok/s | — | 1 |
| Qwen3.6-27B | llama.cpp | IQ2_M | 189.9 tok/s | — | 1 |
| Qwen3.6-27B | llama.cpp | IQ2_XS | 191.2 tok/s | — | 1 |
| Qwen3.6-27B | vLLM | Q3_K_M | 174.2 tok/s | — | 1 |
| Qwen3.6-27B | Ollama | Q6_K | 99.5 tok/s | — | 1 |
| Qwen3.5-35B-A3B | vLLM | IQ4_XS | 115.3 tok/s | — | 1 |
| Qwen3.5-35B-A3B | vLLM | Q4_K_S | 115.9 tok/s | — | 1 |
| Qwen3.5-35B-A3B | Ollama | IQ2_XS | 138 tok/s | — | 1 |
| Qwen3.6-35B-A3B | llama.cpp | IQ3_XS | 121.6 tok/s | — | 1 |
| Qwen3.6-35B-A3B | llama.cpp | Q4_K_S | 111 tok/s | — | 1 |
| Qwen3.6-35B-A3B | Ollama | Q3_K_L | 102.3 tok/s | — | 1 |
| Qwen3.5-4B | llama.cpp | Q5_K_M | 206.3 tok/s | — | 1 |
| Qwen3.5-4B | Ollama | Q3_K_L | 222.8 tok/s | — | 1 |
| Qwen3.5-4B | Ollama | Q4_K_M | 207.8 tok/s | — | 1 |
| Qwen3.5-4B | Strata | IQ4_XS | 330.5 tok/s | — | 1 |
| Qwen3.5-4B | Strata | Q2_K | 349.7 tok/s | — | 1 |
| Qwen3.5-27B | llama.cpp | IQ3_S | 165.6 tok/s | — | 1 |
| Qwen3.5-27B | vLLM | Q3_K_M | 166 tok/s | — | 1 |
| Qwen3.5-27B | Ollama | Q3_K_S | 165.5 tok/s | — | 1 |
| Qwen3.5-27B | Ollama | Q5_K_S | 111.6 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | IQ4_XS | 313.1 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | Q4_K_S | 304.5 tok/s | — | 1 |
| Qwen3.5-0.8B | Strata | Q2_K | 352.9 tok/s | — | 1 |
| Qwen3.5-0.8B | Strata | Q3_K_L | 303.6 tok/s | — | 1 |
| Qwen3.8-9B | vLLM | IQ2_XS | 251.7 tok/s | — | 1 |
| Qwen3.8-9B | Ollama | Q8_0 | 103.2 tok/s | — | 1 |
| Qwen3.8-9B | Strata | IQ2_XS | 353.9 tok/s | — | 1 |
| Qwen3.8-9B | Strata | IQ3_S | 255.9 tok/s | — | 1 |
| Qwen3-8B | llama.cpp | Q2_K | 238 tok/s | — | 1 |
| Qwen3-8B | llama.cpp | Q3_K_L | 189.8 tok/s | — | 1 |
| Qwen3-8B | vLLM | FP16 | 103.6 tok/s | — | 1 |
| Qwen3-8B | vLLM | IQ1_S | 287.1 tok/s | — | 1 |
| Qwen3-8B | Ollama | FP16 | 81 tok/s | — | 1 |
| Qwen3-8B | Ollama | IQ4_XS | 182.1 tok/s | — | 1 |
| Qwen3-1.7B | vLLM | IQ3_S | 318.2 tok/s | — | 1 |
| Qwen3-1.7B | vLLM | Q5_K_S | 322.7 tok/s | — | 1 |
| Qwen3-1.7B | Strata | IQ2_XS | 304.1 tok/s | — | 1 |
| gemma-4-31B-it | llama.cpp | Q3_K_M | 93 tok/s | — | 1 |
| gemma-4-31B-it | llama.cpp | Q8_0 | 53.1 tok/s | — | 1 |
| gemma-4-31B-it | vLLM | Q2_K | 111.5 tok/s | — | 1 |
| gemma-4-31B-it | vLLM | Q8_0 | 54.3 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | IQ2_M | 187.1 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | Q4_K_M | 140.9 tok/s | — | 1 |
| gemma-4-12B-it | vLLM | IQ3_XS | 181.7 tok/s | — | 1 |
| gemma-4-12B-it | Strata | Q2_K | 270.8 tok/s | — | 1 |
| DeepSeek-V4-Flash | llama.cpp | IQ3_S | 3.7 tok/s | — | 1 |
| Qwen3.8-Flash-Next | llama.cpp | IQ3_S | 5.8 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | Q4_K_M | 26 tok/s | — | 1 |