Open-weight models that run on Apple M6 (24GB or 32GB) · 24 GB

Unified memory 24 GB
Bandwidth 170 GB/s
Usable 16.0 GB

Specs: Memory · Mac mini - Technical Specifications - Apple · Bandwidth · Mac mini - Technical Specifications - Apple

Search Precision Context Concurrent requests Minimum speed Architecture Type Maker Sort

Fit	Model	Parameters / active	Precision	Memory	Est. speed	Longest context	Runs as
Runs well	Gemma 4 26B A4B IT Google	25.8B / 3.8B	IQ4_XS	13.9 / 16.0 GB	33–50 tok/s	61K	Unified memory
Runs well	ERNIE-4.5-21B-A3B-Thinking Baidu	21.8B / 3B	Q4_K_M	13.4 / 16.0 GB	37–56 tok/s	55K	Unified memory
Runs well	GPT OSS 20B OpenAI	20.9B / 3.6B	MXFP4 (native)maker’s file	13.6 / 16.0 GB	33–50 tok/s	108K	Unified memory
Runs well	Qwen2.5-Omni-7B Alibaba Qwen	10.7B	Q4_0	6.6 / 16.0 GB	13–20 tok/s	32K	Unified memory
Runs well	GLM-4.6V-Flash Zhipu AI	10.3B	Q4_K_M	6.6 / 16.0 GB	13–19 tok/s	125K	Unified memory
Runs well	Qwen3.5-9B Alibaba Qwen	9.7B	Q4_K_M	6.2 / 16.0 GB	14–21 tok/s	256K	Unified memory
Runs well	Qianfan-VL-8B Baidu	8.8B	Q4_K_M	6.5 / 16.0 GB	13–20 tok/s	32K	Unified memory
Runs well	Granite Switch 4.1 8B IBM	8.8B	IQ4_XS	6.2 / 16.0 GB	14–21 tok/s	68K	Unified memory
Runs well	Qwen3-VL-8B-Instruct Alibaba Qwen	8.8B	Q4_K_Mmaker’s file	6.3 / 16.0 GB	14–20 tok/s	77K	Unified memory
Runs well	Qwen3-VL-8B-Thinking Alibaba Qwen	8.8B	Q4_K_Mmaker’s file	6.3 / 16.0 GB	14–20 tok/s	77K	Unified memory
Runs well	pplx-embed-v2-context-9b Perplexity	8.4B	Q5_K_M	6.3 / 16.0 GB	14–20 tok/s	256K	Unified memory
Runs well	Qwen2.5-VL-7B-Instruct Alibaba Qwen	8.3B	Q5_K_M	6.4 / 16.0 GB	13–20 tok/s	125K	Unified memory
Runs well	Qwen3-8B Alibaba Qwen	8.2B	Q4_K_Mmaker’s file	6.3 / 16.0 GB	14–20 tok/s	128K	Unified memory
Runs well	DeepSeek-R1-Distill-Llama-8B DeepSeek	8B	Q5_K_M	6.8 / 16.0 GB	13–19 tok/s	81K	Unified memory
Runs well	Gemma 4 E4B Google	8B	Q6_K	6.8 / 16.0 GB	13–19 tok/s	125K	Unified memory
Runs well	Ling-3.0-tiny Ant Group	7.9B / 1.3B	Q8_0maker’s file	8.5 / 16.0 GB	53–79 tok/s	128K	Unified memory
Runs well	Qwen2.5-7B-Instruct-1M Alibaba Qwen	7.6B	Q6_K	6.8 / 16.0 GB	13–19 tok/s	177K	Unified memory
Runs well	Qwen2.5-7B-Instruct Alibaba Qwen	7.6B	Q6_Kmaker’s file	6.8 / 16.0 GB	13–19 tok/s	32K	Unified memory
Runs well	DeepSeek-R1-Distill-Qwen-7B DeepSeek	7.6B	Q6_K	6.8 / 16.0 GB	13–19 tok/s	32K	Unified memory
Runs well	Mistral-7B-Instruct-v0.1 Mistral AI	7.2B	Q5_K_M	6.3 / 16.0 GB	14–20 tok/s	32K	Unified memory
Runs well	Gemma 4 E2B Google	5.1B	Q8_0	5.6 / 16.0 GB	15–23 tok/s	125K	Unified memory
Runs well	WeVisOne-VL-4B Tencent	4.8B	Q8_0	6.4 / 16.0 GB	13–20 tok/s	76K	Unified memory
Runs well	Qianfan-OCR Baidu	4.7B	Q8_0	6.3 / 16.0 GB	14–20 tok/s	31K	Unified memory
Runs well	Qwen3.5-4B Alibaba Qwen	4.7B	Q8_0	5.4 / 16.0 GB	16–24 tok/s	256K	Unified memory
Runs well	Qwen3-VL-4B-Instruct Alibaba Qwen	4.4B	Q8_0maker’s file	5.6 / 16.0 GB	15–23 tok/s	82K	Unified memory
Runs well	Qwen3-4B-Instruct-2507 Alibaba Qwen	4B	Q8_0	5.6 / 16.0 GB	16–23 tok/s	82K	Unified memory
Runs well	Qwen3-4B Alibaba Qwen	4B	MLX 8bitmaker’s file	5.5 / 16.0 GB	16–24 tok/s	128K	Unified memory
Runs well	DeepSeek-OCR DeepSeek	3.3B	BF16 (native)maker’s file	7.2 / 16.0 GB	31–46 tok/s	8K	Unified memory
Runs well	Granite-4.0-H-Micro IBM	3.2B	BF16 (native)maker’s file	6.5 / 16.0 GB	13–20 tok/s	128K	Unified memory
Runs well	Qwen3-1.7B Alibaba Qwen	2B	BF16 (native)maker’s file	5.2 / 16.0 GB	17–25 tok/s	40K	Unified memory
Runs well	DeepSeek-R1-Distill-Qwen-1.5B DeepSeek	1.8B	BF16 (native)maker’s file	4.0 / 16.0 GB	22–34 tok/s	125K	Unified memory
Runs well	GLM-OCR Zhipu AI	1.3B	BF16 (native)maker’s file	3.5 / 16.0 GB	27–40 tok/s	128K	Unified memory
Runs well	Qwen3-0.6B Alibaba Qwen	0.8B	BF16 (native)maker’s file	2.8 / 16.0 GB	35–52 tok/s	40K	Unified memory
Runs well	EmbeddingGemma 2 Google	0.7B	BF16 (native)maker’s file	2.0 / 16.0 GB	54–81 tok/s	256K	Unified memory
Runs well	Qwen3-Embedding-0.6B Alibaba Qwen	0.6B	BF16 (native)maker’s file	2.5 / 16.0 GB	40–60 tok/s	32K	Unified memory
Runs well	bge-reranker-v2-m3 BAAI	0.6B	FP32 (native)maker’s file	3.4 / 16.0 GB	28–41 tok/s	8K	Unified memory
Runs well	bge-m3-unsupervised BAAI	0.6B	FP32 (native)maker’s file	3.4 / 16.0 GB	28–41 tok/s	8K	Unified memory
Runs well	bge-large-en-v1.5 BAAI	0.3B	FP32 (native)maker’s file	1.8 / 16.0 GB	61–92 tok/s	512	Unified memory
Runs well	bge-reranker-base BAAI	0.3B	FP32 (native)maker’s file	1.6 / 16.0 GB	75–113 tok/s	514	Unified memory
Runs well	bge-base-en-v1.5 BAAI	0.1B	FP32 (native)maker’s file	0.9 / 16.0 GB	186–279 tok/s	512	Unified memory
Runs well	bge-small-en-v1.5 BAAI	0B	FP32 (native)maker’s file	0.6 / 16.0 GB	595–892 tok/s	512	Unified memory
Runs, slowly	Qwen3.5-27B Alibaba Qwen	27.8B	Q4_0	15.8 / 16.0 GB	5–8 tok/s	11K	Unified memory
Runs, slowly	Qwen3.6-27B Alibaba Qwen	27.8B	Q4_0	15.8 / 16.0 GB	5–8 tok/s	11K	Unified memory
Runs, slowly	Qwen3.8-27B Alibaba Qwen	27.8B	Q4_0	15.8 / 16.0 GB	5–8 tok/s	11K	Unified memory
Runs, slowly	Magistral Small 1.2 Mistral AI	24B	IQ4_XS	13.7 / 16.0 GB	6–9 tok/s	22K	Unified memory
Runs, slowly	Devstral Small 1.0 Mistral AI	23.6B	Q4_0	14.2 / 16.0 GB	6–9 tok/s	19K	Unified memory
Runs, slowly	Devstral Small 1.1 Mistral AI	23.6B	Q4_0	14.2 / 16.0 GB	6–9 tok/s	19K	Unified memory
Runs, slowly	Magistral Small 1.0 Mistral AI	23.6B	Q4_0	14.2 / 16.0 GB	6–9 tok/s	19K	Unified memory
Runs, slowly	Magistral Small 1.1 Mistral AI	23.6B	Q4_0	14.2 / 16.0 GB	6–9 tok/s	19K	Unified memory
Runs, slowly	Qwen2.5-14B-Instruct-1M Alibaba Qwen	14.8B	Q6_K	13.4 / 16.0 GB	6–9 tok/s	22K	Unified memory
Runs, slowly	Qwen2.5-14B-Instruct Alibaba Qwen	14.8B	Q6_Kmaker’s file	13.4 / 16.0 GB	6–9 tok/s	22K	Unified memory
Runs, slowly	DeepSeek-R1-Distill-Qwen-14B DeepSeek	14.8B	Q6_K	13.4 / 16.0 GB	6–9 tok/s	22K	Unified memory
Runs, slowly	Qwen3-14B Alibaba Qwen	14.8B	Q6_Kmaker’s file	13.1 / 16.0 GB	6–9 tok/s	26K	Unified memory
Runs, slowly	Gemma 3 12B IT Google	12.2B	Q6_K	12.6 / 16.0 GB	7–10 tok/s	17K	Unified memory

Estimates from the makers’ Hugging Face repositories (as of 2026-11-06); filtered views aren’t indexed. How it’s estimated