Apple M2 Max

Mac Studio

96 GB RAM · 17 models tested

Avg Score
78 /100
Avg Speed
71.2 tok/s
Models Tested
17
Model Size Score tok/s TTFT Verdict
qwen3-30b-a3b-thinking-2507-claude-4.5-sonnet-high-reasoning-distill-mlx 30B MXFP4 97 79.6 1.7s Excellent
qwen3-30b-a3b-thinking-2507-claude-4.5-sonnet-high-reasoning-distill-mlx 30B MXFP4 95 80.4 1.8s Excellent
qwen3-vl-4b-instruct 4B Q4_K_M 87 72.0 176 ms Excellent
qwen3-vl-4b-instruct 4B Q4_K_M 87 71.2 183 ms Excellent
qwen3-vl-4b-instruct 4B Q4_K_M 87 71.2 156 ms Excellent
qwen3-vl-4b-instruct 4B Q4_K_M 87 71.6 157 ms Excellent
qwen3-vl:4b-instruct 4.4B Q4_K_M 83 90.6 159 ms Excellent
qwen3-vl:4b-instruct 4.4B Q4_K_M 83 91.2 150 ms Excellent
qwen3.6-35b-a3b 35B Q4_K_S 82 55.2 30.0s Not Rec.
qwen3.6-35b-a3b 35B Q4_K_S 82 54.7 30.0s Not Rec.
qwen3.5-35b-a3b-uncensored-hauhaucs-aggressive 35B Q4_K_M 71 59.8 2.4s Good
qwen3.5-35b-a3b-uncensored-hauhaucs-aggressive 35B Q4_K_M 71 59.8 2.4s Good
qwen3-4b-qwen3.6-plus-reasoning-distilled 4B Q4_1 65 99.6 1.6s Good
glm-4.6v-flash Q4_K_S 65 49.9 3.7s Good
qwen3-vl-4b-instruct-q4_k_m.gguf 62 67.8 64 ms Good
qwen3.5-35b-a3b-uncensored-hauhaucs-aggressive 62 67.9 75 ms Good
qwen3-vl-4b-instruct-q4_k_m 62 67.2 68 ms Good