ddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit

mlx-serve 26.10.1 · 12th Gen Intel(R) Core(TM) i5-12600K · 31GB · linux/x64 · 2026-10-15 · ctx113k no-attn-quant

Scores

Coverage100%core tier
Conformance100%210/210 MUST checks
Capability94%strong
Agentic25%2/8 tasks
Fidelity100%36 items

Performance

Decode77.9tok/smedian of 3 · 77.8–78.2
Prefill1990tok/s1989–2000 · 10402 prompt tokens
First token107ms106–107
Speculative1×77.7 predictable vs 78 novel tok/s · none
Prefix cache9.1×858 → 94 ms first token
Concurrency0.51145.9 aggregate vs 70.8 tok/s alone (4 streams)
Sustained load0%77.9 → 77.9 tok/s

Context scaling

Decode tok/s
0661315124.1k8.2k16.4kddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 512: 75.5 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 4.1k: 71.7 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 8.2k: 114.3 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 16.4k: 61 tok/s
First token ms
0483296635124.1k8.2k16.4kddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 512: 354 msddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 4.1k: 2102 msddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 8.2k: 4161 msddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 16.4k: 8403 ms
Prefill tok/s
0116223235124.1k8.2k16.4kddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 512: 1444 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 4.1k: 2020 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 8.2k: 1981 tok/sddalcu/MiMo-V2.6-Distill-Qwen-9B-MLX-Serve-4bit · 16.4k: 1941 tok/s

Capability by category

tool-selection
100%6/6
tool-restraint
100%6/6
tool-args
67%4/6
multiturn
100%3/3
instructions
100%3/3
json-discipline
100%6/6
reasoning
100%3/3
knowledge
100%3/3

Conformance by surface

models
100%4/4
chat
100%70/70
responses
100%66/66
messages
100%63/63
embeddings
100%4/4
completions
100%3/3

Coverage by tier

core
100%9/9
extended
93%14/15
frontier
56%5/9

report.json · measured with llmprobe