Machine

Apple M5 Max · Apple M5 Max #32527cfb

304 runs of 78 models since May 12, 2026. Last run Nov 21, 2026.

ShareRedditX
304
runs
78
models
126
configs
66
configs run 2+ times

Hardware

GPU
Apple M5 Max 64 GB
CPU
Apple M5 Max 18 cores
RAM
64 GB
OS
macOS 26.6.2
Tools
oMLX 283 · Ollama 16 · llama.cpp 4 · LM Studio 1
Client
0.4.99
All Apple M5 Max benchmarks

Models on this machine

One dot per model, at its best config. Up is better answers, right is faster. Time to finish counts every token the model wrote, thinking included, so a model that talks a lot ends up further left.

Best trade-off: quality 60+, and nothing else here is both faster and betterOther models · dot size = number of runs
below 20 tok/s: slower than you read0204060801005102050100tok/s (log scale) →quality →outliers sit at the edges

Medians of each model's best config, 78 models.

Best config per model

sort

Runs with the same model, quant, tool, context, thinking and KV cache count as one config. Numbers are the median of its runs, and each dot is one run on the 0–100 quality scale. ★ marks the best config with at least 2 rated runs.

Qwen3.8-27B-MTPLX-Optimized-Quality
2 runs · 1 config
★ bestoMLX256k ctxthinking oneffort xhighKV none
×2 runs
89.188.9–89.3
steady
34.3 tok/s
32.8–35.8
DaisyRose-27B-oQ8e-mtp
6 runs · 1 config
★ bestoQ8eoMLX256k ctxthinking oneffort xhighKV none
×6 runs
88.884.7–89.1
steady
50.0 tok/s
49.1–52.2
Qwen3.8-27B-oQ4e-mtp
30 runs · 12 configs
★ bestoMLX256k ctxthinking oneffort xhighKV none
×2 runs
87.386.7–88.0
steady
44.2 tok/s
41.2–47.1
Inkling-2-Mini-oQ4e-mtp
6 runs · 1 config
★ bestoQ4eoMLX256k ctxthinking onKV none
×6 runs
87.261.4–88.8
2 outliers
57.6 tok/s
54.1–59.6
2 runs of the best config scored 61.4, 64.6, the other 4 sit between 87.2 and 88.8. The median isn't thrown off by them.
Qwen3.8-27B-oQ8e-fp16-mtp
4 runs · 2 configs
★ bestFP16oMLX64k ctxthinking default
×2 runs
87.285.6–88.8
steady
35.4 tok/s
35.4–35.4
Qwen3.8-27B-oQ8e-mtp
62 runs · 16 configs
★ bestoQ8eoMLX170k ctxthinking oneffort xhighKV none
×7 runs
86.984.5–89.9
steady
51.7 tok/s
46.6–52.3
Qwen3.8-27B-Q8_0
2 runs · 1 config
★ bestQ8_0oMLX256k ctxthinking oneffort xhighKV none
×2 runs
86.585.6–87.3
steady
53.9 tok/s
53.3–54.6
Tiel-Coder-35B-A3B-MLX-oQ6e-MTP
1 run · 1 config
★ bestoQ6eoMLX256k ctxthinking onKV none
86.3
1 rated run
167 tok/s
Swift-Qwen3.8-27b-oQ4e-fp16-mtp
2 runs · 1 config
★ bestFP16oMLX256k ctxthinking oneffort xhighKV none
×2 runs
86.185.5–86.7
steady
46.7 tok/s
44.8–48.6
Hemmingway-1-oQ8e-mtp
4 runs · 1 config
★ bestoQ8eoMLX256k ctxthinking oneffort xhighKV none
×4 runs
86.082.1–87.6
steady1 slow run
31.7 tok/s
17.5–32.0

Compare with another machine

Run history

Show all 304 runs ▾