Machine

Apple M5 Max · Apple M5 Max #5e3bf11b

100 runs of 40 models since Sep 27, 2026. Last run Nov 20, 2026.

ShareRedditX
100
runs
40
models
88
configs
10
configs run 2+ times

Hardware

GPU
Apple M5 Max 64 GB
CPU
Apple M5 Max 18 cores
RAM
64 GB
OS
macOS 27.2.1
Tools
LM Studio 47 · oMLX 20 · Unsloth Studio 14 · splash 6 · llama.cpp 6 · vLLM 4 · OpenAI-compatible 3
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 read4050607080901002050100tok/s (log scale) →quality →outliers sit at the edges

Medians of each model's best config, 40 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.

unsloth/Qwen3.8-27B-GGUF
3 runs · 3 configs
★ bestllama.cpp128k ctxthinking oneffort medium
87.4
1 rated run
29.4 tok/s
qwen/qwen3.8-27b
1 run · 1 config
★ bestLM Studio64k ctxthinking oneffort xhigh
86.4
1 rated run
17.8 tok/s
Muse-Glimmer-30B-4bit
5 runs · 5 configs
★ best4bitoMLX64k ctxthinking defaultKV none
86.3
1 rated run
30.9 tok/s
qwen3.8-27b-splash
10 runs · 7 configs
★ best4bitLM Studio256k ctxthinking on
×2 runs
85.384.6–86.0
steady
73.4 tok/s
71.5–75.3
mlx-community/Qwen3.8-27B-4bit
10 runs · 7 configs
★ bestsplash128k ctxthinking default
×3 runs
85.283.9–85.4
steady
89.3 tok/s
88.5–101
Swift-1.5-Qwen3.8-27b-oQ8e-bf16-mtp-MLX
1 run · 1 config
★ bestoQ8eoMLX64k ctxthinking defaulteffort xhighKV none
85.0
1 rated run
47.0 tok/s
Tiel-Coder-35B-A3B-MLX-oQ4e
2 runs · 2 configs
★ bestoQ4eoMLX128k ctxthinking defaultKV none
84.7
1 rated run
54.5 tok/s
qwen3.8-27b-gsq-rco
1 run · 1 config
★ bestIQ3_SLM Studio128k ctxthinking oneffort xhigh
84.6
1 rated run
43.8 tok/s
muse-glimmer-30b
3 runs · 2 configs
★ best4bitLM Studio128k ctxthinking default
×2 runs
84.082.4–85.6
steady
30.8 tok/s
30.6–31.1
Qwen3.8-27B-oQ4e-mtp
8 runs · 7 configs
★ bestoQ4eoMLX256k ctxthinking oneffort mediumKV none
×2 runs
83.681.7–85.5
steady
29.7 tok/s
29.5–29.9

Compare with another machine

Run history

Show all 100 runs ▾