Apple M4

Mac mini

24 GB RAM · 143 models tested

Avg Score
66 /100
Avg Speed
42.6 tok/s
Models Tested
143
Model Size Score tok/s TTFT Verdict
openai/gpt-oss-20b 20B MXFP4 95 29.3 1.8s Excellent
openai/gpt-oss-20b 20B MXFP4 94 40.7 1.3s Excellent
qwen/qwen3-30b-a3b-2507 30B 4bit 94 44.8 405 ms Excellent
unsloth/gemma-4-26b-a4b-it 26B Q3_K_M 90 24.9 817 ms Excellent
gpt-oss-orchestrator:latest 20.9B MXFP4 88 24.6 4.8s Excellent
gpt-oss:20b 20.9B MXFP4 87 24.3 4.5s Excellent
gpt-oss-safeguard-20b-mlx 20B MXFP4 87 41.4 3.6s Excellent
gpt-oss:20b 20.9B MXFP4 87 26.4 6.7s Excellent
gemma4:26b 25.8B Q4_K_M 85 24.5 1.5s Excellent
qwen2.5:7b 7.6B Q4_K_M 84 22.2 285 ms Excellent
unsloth/gemma-4-26b-a4b-it 26B Q3_K_M 84 16.5 2.2s Excellent
qwen2.5-7b-instruct 7B 4bit 84 24.4 464 ms Excellent
nvidia/nemotron-3-nano 30B 4bit 84 47.9 465 ms Excellent
qwen2.5:14b 14.8B Q4_K_M 83 11.2 539 ms Excellent
gpt-oss:20b 20.9B MXFP4 83 17.1 10.2s Excellent
qwen3:14b 14.8B Q4_K_M 83 11.1 500 ms Excellent
openai/gpt-oss-20b 20B MXFP4 82 31.8 8.4s Excellent
qwen2.5-7b-instruct 7B 4bit 82 22.0 477 ms Excellent
qwen3:14b 14.8B Q4_K_M 82 10.3 513 ms Excellent
qwen3.5-9b 9B 4bit 81 20.4 3.0s Excellent
qwen3.5:9b 9.7B Q4_K_M 80 11.8 523 ms Excellent
gemma3:12b 12.2B Q4_K_M 80 12.7 560 ms Excellent
mlx-community/gemma-3-4b-it-qat-4bit 80 43.2 428 ms Excellent
gemma3:4b 4.3B Q4_K_M 80 34.6 303 ms Excellent
qwen3.5:4b 4.7B Q4_K_M 80 16.6 369 ms Excellent
qwen2.5-coder-7b-instruct-mlx 7B 4bit 79 22.5 529 ms Good
qwen2.5-coder:7b 7.6B Q4_K_M 79 22.1 285 ms Good
yi:6b 6B Q4_0 77 27.6 210 ms Good
qwen2.5:3b 3.1B Q4_K_M 77 47.2 168 ms Good
gemma2:2b 2.6B Q4_0 77 55.1 152 ms Good
deepseek-r1:7b 7.6B Q4_K_M 77 22.2 285 ms Good
internlm2:7b 7.7B Q4_0 77 22.7 250 ms Good
openai/gpt-oss-20b 20B MXFP4 77 30.5 30.0s Not Rec.
cogito:8b 8.0B Q4_K_M 77 21.0 312 ms Good
gemma2:9b 9.2B Q4_0 76 17.5 347 ms Good
ministral-3:8b 8.9B Q4_K_M 76 19.5 328 ms Good
glm4:9b 9.4B Q4_0 76 19.0 290 ms Good
google/gemma-3-4b 4B 4bit 76 39.4 468 ms Good
qwen3.5:27b 27.8B Q4_K_M 75 4.4 1.1s Good
exaone-3.5-2.4b-instruct-mlx 2.4B 8bit 75 37.0 288 ms Good
phi4:14b 14.7B Q4_K_M 75 11.0 532 ms Good
qwen3.5-27b 27B 4bit 75 6.6 4.4s Good
llama3.1:8b 8.0B Q4_K_M 75 20.3 315 ms Good
granite3.1-dense:8b 8.2B Q4_K_M 74 18.8 308 ms Good
minimax-m2.7:cloud 74 0.0 11.4s Not Rec.
liquid/lfm2-24b-a2b 24B 4bit 74 51.8 503 ms Good
mlx-community/Llama-3.2-3B-Instruct-4bit 74 50.6 257 ms Good
llama3.2:3b 3.2B Q4_K_M 74 44.1 178 ms Good
hermes3:8b 8.0B Q4_0 74 22.3 296 ms Good
dolphin3:8b 8.0B Q4_K_M 74 21.0 314 ms Good
liquid/lfm2-24b-a2b 24B 4bit 74 58.9 481 ms Good
mistralai/magistral-small-2509 24B 4bit 74 7.9 1.4s Good
qwen3.5:2b 2.3B Q8_0 73 30.4 215 ms Good
qwen2.5-coder-3b-instruct-mlx 3B 4bit 73 52.3 285 ms Good
mlx-community/Yi-1.5-6B-Chat-4bit 73 29.1 451 ms Good
qwen3:0.6b 751.63M Q4_K_M 73 145.2 1.4s Good
mistral:latest 7.2B Q4_K_M 73 22.5 257 ms Good
mlx-community/gemma-3-1b-it-8bit 72 86.3 253 ms Good
mistral:7b 7.2B Q4_K_M 72 22.0 261 ms Good
qwen3.5-9b 9B Q4_K_S 72 14.4 824 ms Not Rec.
aya-expanse:8b 8.0B Q4_K_M 72 19.7 350 ms Good
codegemma:7b 9B Q4_0 71 18.7 320 ms Good
mistral-nemo:12b 12.2B Q4_0 71 14.6 371 ms Good
ministral-3:3b 3.8B Q4_K_M 71 42.2 189 ms Good
llama3:latest 8.0B Q4_0 71 22.3 295 ms Good
qwen3.5-4b-mlx 4B 4bit 71 23.1 3.3s Good
google/gemma-3-27b 27B 4bit 71 5.9 1.8s Not Rec.
phi4-mini:latest 3.8B Q4_K_M 71 36.2 190 ms Good
qwen2.5-coder-1.5b-instruct-mlx 1.5B 8bit 70 58.4 234 ms Good
qwen2.5:1.5b 1.5B Q4_K_M 70 84.1 119 ms Good
mistral:7b 7.2B Q4_K_M 70 22.5 252 ms Good
qwen3.5-2b-mlx 2B 8bit 70 32.9 3.1s Good
yi-coder:9b 8.8B Q4_0 70 19.4 287 ms Good
deepseek-v2:16b 15.7B Q4_0 69 56.6 239 ms Good
qwen2.5-1.5b-instruct 1.5B 8bit 69 58.1 211 ms Good
mistralai/devstral-small-2-2512 24B 4bit 69 5.7 5.0s Good
lfm2.5-1.2b-instruct-mlx 1.2B 8bit 68 77.3 177 ms Good
granite3.1-dense:2b 2.5B Q4_K_M 68 54.6 121 ms Good
nous-hermes2:latest 11B Q4_0 68 15.9 345 ms Good
gemma3:1b 999.89M Q4_K_M 68 39.4 362 ms Good
gemma3:1b 999.89M Q4_K_M 68 39.5 333 ms Good
cogito:3b 3.6B Q4_K_M 67 45.6 177 ms Good
neural-chat:7b 7B Q4_0 67 22.5 242 ms Good
aya:8b 8.0B F16 66 20.0 300 ms Good
granite-3.3-2b-instruct 2B bf16 66 18.3 541 ms Good
qwen3.5:27b 27.8B Q4_K_M 66 4.0 1.7s Good
deepseek-r1:1.5b 1.8B Q4_K_M 66 85.4 115 ms Good
qwen3.5:27b 27.8B Q4_K_M 66 3.8 1.6s Good
smollm2-1.7b-instruct 1.7B bf16 65 28.5 393 ms Good
starling-lm:7b 7B Q4_0 65 23.8 232 ms Good
qwen3.5:0.8b 873.44M Q8_0 65 48.2 195 ms Good
smollm2:1.7b 1.7B Q8_0 64 51.8 84 ms Good
phi3:3.8b 3.8B Q4_0 64 42.0 142 ms Good
solar:10.7b 11B Q4_0 63 15.8 353 ms Good
starling-lm:7b 7B Q4_0 63 23.7 234 ms Good
phi3:14b 14.0B Q4_0 63 12.7 434 ms Good
codellama:7b 7B Q4_0 63 23.6 297 ms Good
microsoft/phi-4-mini-reasoning 3.8B 4bit 62 40.3 289 ms Good
dolphin-phi:2.7b 3B Q4_0 61 56.1 108 ms Good
qwen3:0.6b 751.63M Q4_K_M 61 144.3 81 ms Good
qwen3.5-0.8b-mlx 0.8B 8bit 60 108.0 2.6s Good
text-embedding-nomic-embed-text-v1.5 Q4_K_M 59 45.3 2.4s Marginal
vicuna:7b 7B Q4_0 59 23.4 242 ms Marginal
vicuna:13b 13B Q4_0 59 13.4 438 ms Marginal
qwen3.5-9b-mlx 9B 4bit 59 14.3 13.4s Marginal
wizardlm2:7b 7B Q4_0 59 23.2 264 ms Marginal
phi:2.7b 3B Q4_0 59 56.6 110 ms Marginal
qwen3.5-9b-mlx 9B 4bit 58 16.7 14.1s Marginal
qwen2.5:0.5b 494.03M Q4_K_M 57 165.0 81 ms Marginal
mlx-community/Llama-3.2-1B-Instruct-4bit 57 126.4 139 ms Marginal
deepseek-coder:6.7b 7B Q4_0 57 24.1 251 ms Marginal
llama2:13b 13B Q4_0 56 13.5 415 ms Marginal
orca-mini:7b 7B Q4_0 55 24.4 219 ms Marginal
phi-3.5-mini-instruct 4bit 55 43.4 359 ms Marginal
qwen3.5-0.8b-mlx 0.8B 4bit 55 68.2 3.3s Marginal
llama2:7b 7B Q4_0 55 25.1 228 ms Marginal
falcon3-1b-instruct 1B 3bit 54 121.8 176 ms Marginal
mlx-community/Nanbeige4.1-3B-8bit 53 25.4 266 ms Marginal
qwen2.5-0.5b-instruct-mlx 0.5B 4bit 52 257.5 99 ms Marginal
smollm2:360m 361.82M F16 52 117.9 43 ms Marginal
smollm2-360m-instruct 360M bf16 52 121.1 106 ms Marginal
mlx-community/stablelm-2-zephyr-1_6b-4bit 50 93.6 166 ms Marginal
qwen3.5:4b 4.7B Q4_K_M 48 18.9 13.0s Marginal
orca-mini:3b 3B Q4_0 47 30.6 126 ms Marginal
qwen3:4b 4.0B Q4_K_M 47 11.3 9.2s Marginal
qwen3:8b 8.2B Q4_K_M 47 18.7 12.1s Marginal
stablelm2:1.6b 2B Q4_0 45 93.8 94 ms Marginal
qwen2.5-math-1.5b-instruct 1.5B 4bit 45 91.7 205 ms Marginal
mlx-community/quantized-gemma-2b-it 45 44.4 336 ms Marginal
tinyllama:1.1b 1B Q4_0 43 121.0 82 ms Marginal
tinyllama 43 133.4 61 ms Marginal
gemma-3-270m-it-qat-mlx 270M 4bit 43 344.2 173 ms Marginal
smollm2:135m 134.52M F16 41 241.0 39 ms Marginal
starcoder2:3b 3B Q4_0 39 49.4 155 ms Not Rec.
phi-3-mini-128k-instruct 4bit 38 44.0 385 ms Not Rec.
deepseek-r1-distill-qwen-14b-mlx 14B 5bit 36 10.3 20.7s Not Rec.
starcoder2:7b 7B Q4_0 35 23.8 251 ms Not Rec.
qwen3.5-27b-claude-4.6-opus-distilled-mlx 27B 4bit 34 6.1 22.9s Not Rec.
qwen3.5:latest 9.7B Q4_K_M 34 13.3 19.8s Not Rec.
gemma4:26b 25.8B Q4_K_M 33 20.0 30.0s Not Rec.
lmstudio-community/Phi-4-reasoning-plus-MLX-4bit 33 11.5 670 ms Not Rec.
deepseek-r1:14b 14.8B Q4_K_M 22 11.4 30.0s Not Rec.
qwen3.5:27b 27.8B Q4_K_M 4 3.5 52.0s Not Rec.