Apple M3 Max
LaptopApple unified-memory SoC; this entry is the 36 GB configuration (14-core CPU / 30-core GPU) with 300 GB/s of memory bandwidth. The 16-core CPU / 40-core GPU M3 Max runs at 400 GB/s and supports up to 128 GB of unified memory. Apple does not publish a TDP, so that column is left empty.
41
Run records
14
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.
- llama.cpp · IQ3_XS Decode 96.32 · Prefill 353.29 ·
- Ollama · IQ2_XS Decode 109.19 · Prefill 420.91 ·
- Ollama · IQ2_M Decode 78.91 · Prefill 300.73 ·
- Strata · IQ2_M Decode 121.51 · Prefill 396.4 ·
- llama.cpp · IQ2_XS Decode 89.41 · Prefill 406.89 ·
- vLLM · IQ1_S Decode 102.56 · Prefill 411.25 ·
- llama.cpp · IQ2_M Decode 188.61 · Prefill 924.46 ·
- llama.cpp · Q6_K Decode 107.73 · Prefill 580.34 ·
- Ollama · Q4_K_S Decode 130.85 · Prefill 560.14 ·
- Strata · Q3_K_M Decode 237.4 · Prefill 1147.78 ·
- Strata · Q6_K Decode 145.1 · Prefill 751.14 ·
- Strata · Q2_K Decode 150.5 · Prefill 804.01 ·
- llama.cpp · IQ2_XS Decode 369.47 · Prefill 2026.25 ·
- llama.cpp · IQ3_XS Decode 313.86 · Prefill 1064.12 ·
- Ollama · Q3_K_M Decode 286.62 · Prefill 1285.71 ·
- Strata · Q8_0 Decode 237.48 · Prefill 1069.27 ·
- llama.cpp · IQ2_M Decode 136.47 · Prefill 438.69 ·
- llama.cpp · IQ3_S Decode 126.59 · Prefill 664.89 ·
- Strata · Q2_K Decode 180.32 · Prefill 900.55 ·
- llama.cpp · FP16 Decode 58.36 · Prefill 290.84 ·
- llama.cpp · IQ2_M Decode 137.92 · Prefill 603.9 ·
- vLLM · FP16 Decode 57.8 · Prefill 255.66 ·
- vLLM · IQ2_M Decode 145.31 · Prefill 613.3 ·
- Ollama · IQ2_M Decode 129.83 · Prefill 636.38 ·
- vLLM · IQ2_XS Decode 309.15 · Prefill 1109.54 ·
- vLLM · Q3_K_L Decode 226.47 · Prefill 1116.37 ·
- Ollama · FP16 Decode 96.85 · Prefill 510.37 ·
- Ollama · IQ4_XS Decode 188.36 · Prefill 704.89 ·
- Strata · Q5_K_S Decode 239.13 · Prefill 849.25 ·
- llama.cpp · Q2_K Decode 64.47 · Prefill 330.47 ·
- Ollama · Q2_K Decode 65.74 · Prefill 261.65 ·
- llama.cpp · Q4_K_M Decode 85.7 · Prefill 345.14 ·
- llama.cpp · Q5_K_S Decode 77.34 · Prefill 357.59 ·
- Ollama · IQ3_S Decode 94.58 · Prefill 396.89 ·
- llama.cpp · IQ2_XS Decode 0.9 · Prefill 6.8 ·
- llama.cpp · IQ3_S Decode 0.8 · Prefill 6 ·
- llama.cpp · IQ2_XS Decode 0.9 · Prefill 6.8 ·
- llama.cpp · IQ3_S Decode 0.8 · Prefill 6 ·
- llama.cpp · IQ2_XS Decode 25.7 · Prefill 192.8 ·
- llama.cpp · Q5_K_M Decode 11.7 · Prefill 87.8 ·
- llama.cpp · Q4_K_M Decode 14.3 · Prefill 107.3 ·
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
- M3 Max
- VRAM
- 36 GB
- Memory bandwidth
- 300 GB/s
- Release date
- 2023-10-30
Run records
41 configurations
| Model | Framework | Quantization | Decode | Gen time | Samples |
|---|---|---|---|---|---|
| Qwen3.8-27B | llama.cpp | IQ3_XS | 96.3 tok/s | — | 1 |
| Qwen3.6-27B | Ollama | IQ2_XS | 109.2 tok/s | — | 1 |
| Qwen3.5-35B-A3B | Ollama | IQ2_M | 78.9 tok/s | — | 1 |
| Qwen3.5-35B-A3B | Strata | IQ2_M | 121.5 tok/s | — | 1 |
| Qwen3.6-35B-A3B | llama.cpp | IQ2_XS | 89.4 tok/s | — | 1 |
| Qwen3.6-35B-A3B | vLLM | IQ1_S | 102.6 tok/s | — | 1 |
| Qwen3.5-4B | llama.cpp | IQ2_M | 188.6 tok/s | — | 1 |
| Qwen3.5-4B | llama.cpp | Q6_K | 107.7 tok/s | — | 1 |
| Qwen3.5-4B | Ollama | Q4_K_S | 130.8 tok/s | — | 1 |
| Qwen3.5-4B | Strata | Q3_K_M | 237.4 tok/s | — | 1 |
| Qwen3.5-4B | Strata | Q6_K | 145.1 tok/s | — | 1 |
| Qwen3.5-27B | Strata | Q2_K | 150.5 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | IQ2_XS | 369.5 tok/s | — | 1 |
| Qwen3.5-0.8B | llama.cpp | IQ3_XS | 313.9 tok/s | — | 1 |
| Qwen3.5-0.8B | Ollama | Q3_K_M | 286.6 tok/s | — | 1 |
| Qwen3.5-0.8B | Strata | Q8_0 | 237.5 tok/s | — | 1 |
| Qwen3.8-9B | llama.cpp | IQ2_M | 136.5 tok/s | — | 1 |
| Qwen3.8-9B | llama.cpp | IQ3_S | 126.6 tok/s | — | 1 |
| Qwen3.8-9B | Strata | Q2_K | 180.3 tok/s | — | 1 |
| Qwen3-8B | llama.cpp | FP16 | 58.4 tok/s | — | 1 |
| Qwen3-8B | llama.cpp | IQ2_M | 137.9 tok/s | — | 1 |
| Qwen3-8B | vLLM | FP16 | 57.8 tok/s | — | 1 |
| Qwen3-8B | vLLM | IQ2_M | 145.3 tok/s | — | 1 |
| Qwen3-8B | Ollama | IQ2_M | 129.8 tok/s | — | 1 |
| Qwen3-1.7B | vLLM | IQ2_XS | 309.1 tok/s | — | 1 |
| Qwen3-1.7B | vLLM | Q3_K_L | 226.5 tok/s | — | 1 |
| Qwen3-1.7B | Ollama | FP16 | 96.8 tok/s | — | 1 |
| Qwen3-1.7B | Ollama | IQ4_XS | 188.4 tok/s | — | 1 |
| Qwen3-1.7B | Strata | Q5_K_S | 239.1 tok/s | — | 1 |
| gemma-4-31B-it | llama.cpp | Q2_K | 64.5 tok/s | — | 1 |
| gemma-4-31B-it | Ollama | Q2_K | 65.7 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | Q4_K_M | 85.7 tok/s | — | 1 |
| gemma-4-12B-it | llama.cpp | Q5_K_S | 77.3 tok/s | — | 1 |
| gemma-4-12B-it | Ollama | IQ3_S | 94.6 tok/s | — | 1 |
| GLM-5.3 | llama.cpp | IQ2_XS | 0.9 tok/s | — | 1 |
| GLM-5.3 | llama.cpp | IQ3_S | 0.8 tok/s | — | 1 |
| DeepSeek-V4.1-Flash | llama.cpp | IQ2_XS | 0.9 tok/s | — | 1 |
| DeepSeek-V4.1-Flash | llama.cpp | IQ3_S | 0.8 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | IQ2_XS | 25.7 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | Q5_K_M | 11.7 tok/s | — | 1 |
| Qwen3.8-27B | llama.cpp | Q4_K_M | 14.3 tok/s | — | 1 |