← All Mac mini models

Mac mini M2

The M2 Mac mini runs local models at 100 GB/s of memory bandwidth with 8 to 24 GB of unified memory. On Apple Silicon that memory is shared with the GPU, so the whole pool is available for weights: at 24 GB you can hold roughly a 31B dense model at Q4. A base M chip is the narrow end of the memory bus. It runs small models pleasantly and stops hard at the memory ceiling, which is the constraint you will hit first.

Apple no longer sells this configuration new. It stays fully evaluated here because the used market is where most of its local AI value now sits.

Specifications

ChipApple M2
CPU cores8
GPU cores10
Unified memory8, 16, or 24 GB
Memory bandwidth100 GB/s
Neural Engine15.8 TOPS
Released2022
AvailabilityUsed market

Memory bandwidth is faster than 6% of the Apple Silicon chips shipped in a Mac, against a 819 GB/s peak.

Its memory ceiling is above 6% of them, against a 512 GB peak.

Pick your configuration

Every option Apple sells with this chip. The model list below recomputes against the one you pick.

Unified memory

What each memory option runs

Unified memory is the ceiling and it is soldered, so this is the decision you cannot revisit.

8 GB unified memory

7 GB usable for weights · 100 GB/s

2,863 of 3,641 models fit, and 1,869 of them run with headroom rather than as a squeeze.

Largest model at Q4
moondream3 preview · 9.27B
Best all-round pick
Qwen3.5 0.8B · Q8_0 · ~60 tok/s

16 GB unified memory

14 GB usable for weights · 100 GB/s

3,128 of 3,641 models fit, and 2,907 of them run with headroom rather than as a squeeze.

Largest model at Q4
DeepSeek V4 Flash JANGTQ2 · 20.2B
Best all-round pick
Qwen3.5 0.8B · Q8_0 · ~60 tok/s

24 GB unified memory

21 GB usable for weights · 100 GB/s

3,352 of 3,641 models fit, and 3,007 of them run with headroom rather than as a squeeze.

Largest model at Q4
Qwen3 VL 30B A3B Instruct · 31.07B
Best all-round pick
Qwen3.5 0.8B · Q8_0 · ~60 tok/s

What a 8 GB M2 Mac mini can run

Every model in the database against this exact configuration, at 100 GB/s. Ratings and speeds are the same numbers the model pages show.

Showing 3641 of 3641 models

Qwen3.5 0.8B

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
1.5 GB18% of RAM~60 tok/sEstimated0.87B params
Run with ToolPiper
Qwen3.5 0.8B Base

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
1.5 GB18% of RAM~60 tok/sEstimated0.87B params
Run with ToolPiper
LFM2 ColBERT 350M

General · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2.5 1.2B Thinking

Reasoning · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2 350M

General · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2 700M

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.3 GB17% of RAM~71 tok/sEstimated0.74B params
Run with ToolPiper
LFM2 350M Extract

General · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2 350M PII Extract JP

General · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2 350M ENJP MT

General · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2 350M Math

Reasoning · Liquid AI · 2025-11-28

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
LFM2 1.2B

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2.5 1.2B Base

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2 1.2B Extract

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2 1.2B Tool

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2 1.2B RAG

General · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
Qwen3.5 2B

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
3.0 GB38% of RAM~23 tok/sEstimated2.27B params
Run with ToolPiper
Qwen3.5 2B Base

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
3.0 GB38% of RAM~23 tok/sEstimated2.27B params
Run with ToolPiper
LFM2 VL 450M

Multimodal · Liquid AI · 2025-11-28

Q8_0Excellent
1.0 GB13% of RAM~116 tok/sEstimated0.45B params
Run with ToolPiper
LFM2.5 1.2B Instruct

Chat · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2.5 1.2B JP

Chat · Liquid AI · 2025-11-28

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
LFM2.5 VL 1.6B

Multimodal · Liquid AI · 2025-11-28

Q8_0Excellent
2.3 GB29% of RAM~33 tok/sEstimated1.6B params
Run with ToolPiper
LFM2 VL 1.6B

Multimodal · Liquid AI · 2025-11-28

Q8_0Excellent
2.3 GB28% of RAM~33 tok/sEstimated1.58B params
Run with ToolPiper
LFM2.5 Audio 1.5B

General · Liquid AI · 2025-11-28

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
LFM2 Audio 1.5B

General · Liquid AI · 2025-11-28

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
Jolia

General · raidium · 2026-06-15

Q8_0Excellent
0.5 GB7% of RAM~2,619 tok/sEstimated0.02B params
Run with ToolPiper
embeddinggemma GTAIDE 300m 2605

Embedding · taide · 2026-06-12

Q8_0Excellent
0.8 GB10% of RAM~175 tok/sEstimated0.3B params
Run with ToolPiper
Qwen3 0.6B

General · Alibaba · 2025-04-27

Q8_0Excellent
1.3 GB17% of RAM~70 tok/sEstimated0.75B params
Run with ToolPiper
Qwen2 0.5B

General · Alibaba

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
surya ocr 2

Multimodal · datalab-to

Q8_0Excellent
1.3 GB16% of RAM~76 tok/sEstimated0.69B params
Run with ToolPiper
distil lfm25 shellper

General · distil-labs

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
reformer crime and punishment

General · Google

Q8_0Excellent
0.5 GB6% of RAM~5,238 tok/sEstimated0.01B params
Run with ToolPiper
gpt oss safeguard 120b

General · openai

Q8_0Excellent
3.1 GB39% of RAMBenchmark needed2.37B params
Run with ToolPiper
PaddleOCR VL 1.5

General · paddlepaddle

Q8_0Excellent
1.6 GB20% of RAM~55 tok/sEstimated0.96B params
Run with ToolPiper
LFM2.5 350M Base

General · Liquid AI

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
ERNIE 4.5 0.3B PT

General · baidu

Q8_0Excellent
0.9 GB11% of RAM~146 tok/sEstimated0.36B params
Run with ToolPiper
Tashkeel 700M

General · etherll

Q8_0Excellent
1.3 GB17% of RAM~71 tok/sEstimated0.74B params
Run with ToolPiper
EXAONE 4.0 1.2B

General · LG AI · 2025-07-15

Q8_0Excellent
1.8 GB23% of RAM~44 tok/sEstimated1.2B params
Run with ToolPiper
OTel LLM 20B IT

General · farbodtavakkoli

Q8_0Excellent
2.5 GB31% of RAMBenchmark needed1.77B params
Run with ToolPiper
FastVLM 0.5B

General · kamilamila

Q8_0Excellent
1.2 GB15% of RAM~84 tok/sEstimated0.62B params
Run with ToolPiper
PaddleOCR VL

General · paddlepaddle

Q8_0Excellent
1.6 GB20% of RAM~55 tok/sEstimated0.96B params
Run with ToolPiper
LFM2.5 350M home assistant sft

General · ordenwills

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
granite 3.1 1b a400m base

General · ibm-granite

Q8_0Excellent
2.0 GB25% of RAMBenchmark needed1.33B params
Run with ToolPiper
Qwen3.5 0.8B Claude 4.6 Opus Reasoning Distilled

Reasoning · jackrong

Q8_0Excellent
1.5 GB18% of RAM~60 tok/sEstimated0.87B params
Run with ToolPiper
Qwen3.5 0.8B Coder Calude Full

Coding · rahul7star

Q8_0Excellent
1.5 GB18% of RAM~60 tok/sEstimated0.87B params
Run with ToolPiper
Dayhoff 170m GR

General · Microsoft

Q8_0Excellent
0.7 GB9% of RAMBenchmark needed0.17B params
Run with ToolPiper
Dayhoff 170m UR90

General · Microsoft

Q8_0Excellent
0.7 GB9% of RAMBenchmark needed0.17B params
Run with ToolPiper
Hunyuan 0.5B Pretrain

General · tencent

Q8_0Excellent
1.1 GB14% of RAM~97 tok/sEstimated0.54B params
Run with ToolPiper
GLM 0.5B

General · primeintellect

Q8_0Excellent
1.1 GB14% of RAMBenchmark needed0.54B params
Run with ToolPiper
GLM OCR

Multimodal · zai-org

Q8_0Excellent
2.0 GB25% of RAM~39 tok/sEstimated1.33B params
Run with ToolPiper
MiniCPM V 4.6

Multimodal · openbmb

Q8_0Excellent
2.0 GB24% of RAM~40 tok/sEstimated1.3B params
Run with ToolPiper
DeepSeek R1 Distill Qwen 1.5B

Reasoning · DeepSeek

Q8_0Excellent
2.5 GB31% of RAM~29 tok/sEstimated1.78B params
Run with ToolPiper
MiniCPM5 1B

General · openbmb

Q8_0Excellent
1.7 GB21% of RAM~49 tok/sEstimated1.08B params
Run with ToolPiper
Ilama 3.2 1B

General · hmellor

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
OTel LLM 1.2B IT

General · farbodtavakkoli

Q8_0Excellent
1.8 GB23% of RAM~43 tok/sEstimated1.21B params
Run with ToolPiper
MiniCPM5 1B SFT

General · openbmb

Q8_0Excellent
1.7 GB21% of RAM~49 tok/sEstimated1.08B params
Run with ToolPiper
Orpo Llama 3.2 1B 15k

General · adamlucek

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
llama 3.2 1b code instruct

Coding · shahriarferdoush

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
Llama 3.2 1B Aegis SFT DPO

General · ahczhg

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
CyberXP_Agent_Llama_3.2_1B

General · abaryan

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
llama3.2_1b_2025_uncensored_v2

General · carsenk

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
OpenReasoning Nemotron 1.5B

Reasoning · nvidia

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.54B params
Run with ToolPiper
Qwen3 ASR 0.6B

General · bezzam

Q8_0Excellent
1.4 GB17% of RAM~67 tok/sEstimated0.78B params
Run with ToolPiper
granite 3.1 3b a800m base

General · ibm-granite

Q8_0Excellent
4.2 GB52% of RAMBenchmark needed3.3B params
Run with ToolPiper
ODIN C1

General · skis-ai-research

Q8_0Excellent
1.8 GB23% of RAM~45 tok/sEstimated1.17B params
Run with ToolPiper
llama3.2_1B_vl

General · thkim0305

Q8_0Excellent
1.6 GB20% of RAM~52 tok/sEstimated1B params
Run with ToolPiper
Qwen3 Coder Next DFlash

Coding · z-lab

Q8_0Excellent
1.7 GB21% of RAM~49 tok/sEstimated1.08B params
Run with ToolPiper
LLAMA 3.2 1B OpenHermes2.5

General · artificialguybr

Q8_0Excellent
1.9 GB24% of RAM~42 tok/sEstimated1.24B params
Run with ToolPiper
Qwen2.5 0.5B

General · Alibaba

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Qwen2.5 1.5B

General · Alibaba

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.54B params
Run with ToolPiper
SmolLM3 3B

Reasoning · HuggingFace · 2025-07-08

Q8_0Excellent
3.8 GB48% of RAM~17 tok/sEstimated3B params
Run with ToolPiper
Qwen3 0.6B Base

General · Alibaba

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
harrier oss v1 0.6b

General · Microsoft

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
InternVL3 1B hf

Multimodal · opengvlab

Q8_0Excellent
1.5 GB19% of RAM~56 tok/sEstimated0.94B params
Run with ToolPiper
Qwen3Guard Gen 0.6B

General · Alibaba

Q8_0Excellent
1.3 GB17% of RAM~70 tok/sEstimated0.75B params
Run with ToolPiper
voyage 4 nano

General · voyageai

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
harrier oss v1 270m

General · Microsoft

Q8_0Excellent
0.8 GB10% of RAM~194 tok/sEstimated0.27B params
Run with ToolPiper
OTel LLM 0.6B IT

General · farbodtavakkoli

Q8_0Excellent
1.0 GB13% of RAM~116 tok/sEstimated0.45B params
Run with ToolPiper
OTel LLM 270M IT

General · farbodtavakkoli

Q8_0Excellent
0.8 GB10% of RAM~194 tok/sEstimated0.27B params
Run with ToolPiper
ReaderLM v2

General · jinaai

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.54B params
Run with ToolPiper
pii extractor gemma 3 270m it

General · jakobhuss

Q8_0Excellent
0.8 GB10% of RAM~194 tok/sEstimated0.27B params
Run with ToolPiper
llava interleave qwen 0.5b hf

General · llava-hf

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
plamo 2 1b

General · pfnet

Q8_0Excellent
1.9 GB24% of RAM~41 tok/sEstimated1.29B params
Run with ToolPiper
PARD Llama 3.2 1B

General · amd

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
MiniCPM4 0.5B

General · openbmb

Q8_0Excellent
1.0 GB12% of RAM~122 tok/sEstimated0.43B params
Run with ToolPiper
Kurtis EON1 Hybrid 0.7B v0.1.1

General · mrs83

Q8_0Excellent
1.2 GB16% of RAM~78 tok/sEstimated0.67B params
Run with ToolPiper
deepseek coder 1.3b base

Coding · DeepSeek

Q8_0Excellent
1.6 GB20% of RAM~52 tok/sEstimated1.01B params
Run with ToolPiper
granite 4.0 350m base

General · ibm-granite

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
NuExtract tiny

General · numind

Q8_0Excellent
1.0 GB13% of RAM~114 tok/sEstimated0.46B params
Run with ToolPiper
PolyGuard Qwen Smol

General · toxicityprompts

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Heron NVILA Lite 1B hf

General · turing-motors

Q8_0Excellent
1.5 GB19% of RAM~58 tok/sEstimated0.91B params
Run with ToolPiper
granite 4.0 350m

General · ibm-granite

Q8_0Excellent
0.9 GB11% of RAM~150 tok/sEstimated0.35B params
Run with ToolPiper
transcription cleanup llama3.2 1b

General · getonit

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
SpatialLM1.1 Qwen 0.5B Structured3D SFT

General · ysmao

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
SEX_ROLEPLAY 3.2 1B

General · novaciano

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
llava onevision qwen2 0.5b ov

General · lmms-lab

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
Qwen3 0.3B distil

General · tabularisai

Q8_0Excellent
0.9 GB12% of RAM~138 tok/sEstimated0.38B params
Run with ToolPiper
NuExtract 1.5 tiny

General · numind

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
JiRackTernary_1b

General · kgrabko

Q8_0Excellent
2.2 GB27% of RAM~35 tok/sEstimated1.5B params
Run with ToolPiper
Recurrent Llama 3.2 train recurrence 32

General · smcleish

Q8_0Excellent
2.1 GB26% of RAM~38 tok/sEstimated1.39B params
Run with ToolPiper
TinyMixtral 4x248M MoE

General · isotonic

Q8_0Excellent
1.3 GB16% of RAMBenchmark needed0.7B params
Run with ToolPiper
Nemotron Flash 1B

General · nvidia

Q8_0Excellent
1.6 GB20% of RAMBenchmark needed0.97B params
Run with ToolPiper
SpatialLM1.1 Qwen 0.5B

General · manycore-research

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
Qwen3 0.6B diffusion mdlm v0.1

General · dllm-hub

Q8_0Excellent
1.3 GB17% of RAM~70 tok/sEstimated0.75B params
Run with ToolPiper
MiniCPM V 4.6 Thinking

General · openbmb

Q8_0Excellent
2.0 GB24% of RAM~40 tok/sEstimated1.3B params
Run with ToolPiper
granite 4.0 h 1b

General · ibm-granite

Q8_0Excellent
2.1 GB27% of RAM~36 tok/sEstimated1.46B params
Run with ToolPiper
MiniCPM V 4.6 Abliterated AND Disinhibited

General · treadon

Q8_0Excellent
2.0 GB24% of RAM~40 tok/sEstimated1.3B params
Run with ToolPiper
Huihui MoE 0.8B 2E

General · huihui-ai

Q8_0Excellent
1.5 GB18% of RAMBenchmark needed0.86B params
Run with ToolPiper
PARD Qwen3 0.6B

General · amd

Q8_0Excellent
1.3 GB17% of RAM~70 tok/sEstimated0.75B params
Run with ToolPiper
Lucy 128k

General · menlo

Q8_0Excellent
2.4 GB30% of RAM~30 tok/sEstimated1.72B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune hier col mask moe sparse 4e 2k sp0.9_20260418_063208

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune col mask moe sparse 4e 2k sp0.9_20260418_062617

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune col mask moe sparse 4e 2k sp0.9_20260418_062427

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
CeluneNorm 0.6B v1.1

General · lunahr

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
Falcon H1 1.5B Base

General · TII

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.55B params
Run with ToolPiper
granite 4.0 h 350m base

General · ibm-granite

Q8_0Excellent
0.9 GB11% of RAM~154 tok/sEstimated0.34B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune col mask moe sparse 4e 2k sp0.7_20260416_182923

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
CeluneNorm 0.6B v1.2

General · lunahr

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
evolai qwen2.5 1.5b

General · roystar

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.54B params
Run with ToolPiper
Qwen3 0.6B diffusion bd3lm v0.1

General · dllm-hub

Q8_0Excellent
1.3 GB17% of RAM~70 tok/sEstimated0.75B params
Run with ToolPiper
Falcon H1 1.5B Deep Base

General · TII

Q8_0Excellent
2.2 GB28% of RAM~34 tok/sEstimated1.55B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune dyn col mask moe sparse 4e 2k sp0.9 r32_20260422_002811

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune latent sparse moe 4e 1k_20260413_214436

General · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
canine c

General · Google

Q8_0Excellent
0.6 GB8% of RAM~403 tok/sEstimated0.13B params
Run with ToolPiper
HunyuanOCR

Multimodal · tencent

Q8_0Excellent
1.6 GB20% of RAM~52 tok/sEstimated1B params
Run with ToolPiper
GOT OCR2_0

Multimodal · stepfun-ai

Q8_0Excellent
1.3 GB16% of RAM~73 tok/sEstimated0.72B params
Run with ToolPiper
VideoLLaMA3 2B Image HF

Multimodal · lkhl

Q8_0Excellent
2.7 GB34% of RAM~27 tok/sEstimated1.96B params
Run with ToolPiper
Falcon H1 0.5B Base

General · TII

Q8_0Excellent
1.1 GB14% of RAM~101 tok/sEstimated0.52B params
Run with ToolPiper
internlm2_5 step prover critic

General · internlm

Q8_0Excellent
2.1 GB26% of RAM~37 tok/sEstimated1.4B params
Run with ToolPiper
canine s

General · Google

Q8_0Excellent
0.6 GB8% of RAM~403 tok/sEstimated0.13B params
Run with ToolPiper
OTel LLM 1B IT

General · farbodtavakkoli

Q8_0Excellent
1.6 GB20% of RAM~52 tok/sEstimated1B params
Run with ToolPiper
typhoon ocr1.5 2b

Reasoning · typhoon-ai

Q8_0Excellent
2.9 GB36% of RAM~25 tok/sEstimated2.13B params
Run with ToolPiper
deepseek coder 1.3b instruct

Coding · DeepSeek

Q8_0Excellent
2.0 GB25% of RAM~39 tok/sEstimated1.35B params
Run with ToolPiper
Qwen2.5 Coder 0.5B

Coding · Alibaba

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
F2LLM v2 80M

Coding · codefuse-ai

Q8_0Excellent
0.6 GB7% of RAM~655 tok/sEstimated0.08B params
Run with ToolPiper
Ouro 1.4B Thinking

General · bytedance

Q8_0Excellent
2.1 GB26% of RAM~37 tok/sEstimated1.43B params
Run with ToolPiper
nanoLLaVA

General · qnguyen3

Q8_0Excellent
1.7 GB21% of RAM~50 tok/sEstimated1.05B params
Run with ToolPiper
F2LLM v2 0.6B

Coding · codefuse-ai

Q8_0Excellent
1.2 GB15% of RAM~87 tok/sEstimated0.6B params
Run with ToolPiper
InternVL3_5 1B HF

General · opengvlab

Q8_0Excellent
1.7 GB21% of RAM~49 tok/sEstimated1.06B params
Run with ToolPiper
DeepScaleR 1.5B Preview

General · agentica-org

Q8_0Excellent
2.5 GB31% of RAM~29 tok/sEstimated1.78B params
Run with ToolPiper
Qwen2.5 Coder 0.5B Instruct Gensyn Swarm slithering_hairy_lemur

Coding · candan77

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Qwen2.5 Coder 0.5B Instruct Gensyn Swarm leggy_lithe_mallard

Coding · tommymir4444

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Ovis2 1B hf

General · thisisiron

Q8_0Excellent
1.8 GB22% of RAM~46 tok/sEstimated1.13B params
Run with ToolPiper
Qwen2.5 Coder 0.5B Instruct Gensyn Swarm leggy_jagged_hawk

Coding · holten51

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Qwen2.5 Coder 0.5B Instruct Gensyn Swarm bipedal_extinct_owl

Coding · baebii

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
Qwen2.5 Coder 0.5B Instruct Gensyn Swarm fishy_camouflaged_flea

Coding · bedeviler

Q8_0Excellent
1.0 GB13% of RAM~107 tok/sEstimated0.49B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune dyn col mask moe sparse 4e 2k sp0.9 r16_20260418_062801

Reasoning · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune dyn hier col mask moe sparse 4e 2k sp0.9 r16_20260418_063238

Reasoning · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
llavaqwen2.5 0.5b finetune dyn col mask moe sparse 4e 2k sp0.9 r16_20260420_180515

Reasoning · kkhya

Q8_0Excellent
1.1 GB13% of RAM~105 tok/sEstimated0.5B params
Run with ToolPiper
VibeThinker 1.5B

General · weiboai

Q8_0Excellent
2.5 GB31% of RAM~29 tok/sEstimated1.78B params
Run with ToolPiper
gemma 3 1b it fixed

General · pjmixers-dev

Q8_0Excellent
1.6 GB20% of RAM~52 tok/sEstimated1B params
Run with ToolPiper

Measured on the M2

Nobody has submitted a benchmark on the M2 yet, so every speed on this page is the formula estimate rather than a measured run. The estimate is bandwidth-driven and calibrated against chips that do have data, which makes it a good guide and not a promise.

ToolPiper contributes a result anonymously when you run the benchmark, and the leaderboard shows every chip that already has one.

Where to go from here

What each step actually changes for local models, rather than which one is newer.

One tier up

Mac mini M2 Pro

2x the memory bandwidth, up to 32 GB instead of 24 GB

Newer generation

Mac mini M4

1.2x the memory bandwidth, up to 32 GB instead of 24 GB

Used market alternative

Mac mini M1

32% less memory bandwidth, 16 GB ceiling instead of 24 GB

Same chip, other Mac

MacBook Air M2

The same chip in a different Mac

Why unified memory is the number that matters

On a PC the model has to fit in GPU VRAM, which is a separate pool from system RAM and usually the smaller of the two. Apple Silicon has one pool. The M2's 100 GB/s bus is shared by CPU, GPU, and Neural Engine, so a 24 GB machine can hand almost all of that to a model with no copy across a bus.

Apple stopped selling this one, which is exactly why it is interesting. The mini is the cheapest way onto this chip, and it runs headless on a shelf perfectly well. Nothing about local inference needs the display attached. A used M2 at 24 GB still gives you 100 GB/s and a hard 31B ceiling, and neither number degrades with age the way a battery does.

Common questions

Can the M2 Mac mini run a 70B model?

No. A 70B model at Q4_K_M needs about 46 GB, and the largest M2 Mac mini tops out at 24 GB, which leaves about 21 GB for weights. The practical ceiling on this machine is around 31B parameters at Q4.

How much unified memory should I get with the M2 Mac mini?

Memory is the only spec that changes what you can run at all. 8 GB holds about a 9B model at Q4; 24 GB holds about 31B. It is soldered, so this is a one-time decision, and it is the upgrade worth paying for before core count.

How fast are local LLMs on the M2?

Token generation is bandwidth-bound, so M2 throughput scales with its 100 GB/s memory bus. Divide bandwidth by the size of the weights actually read per token to get the ceiling, then expect roughly half of that in practice. A 7B model at Q4 reads about 4 GB per token pass, so the M2 lands in the tens of tokens per second and a 70B model lands in the single digits.

Is a used M2 Mac mini still worth buying for local AI?

For inference, the specs that matter do not age: 100 GB/s and up to 24 GB of unified memory are the same numbers today as they were in 2022. A used M2 at the top memory option usually beats a new base-tier machine at the same price on both. Check the battery and the display, not the silicon.

Run these models on your Mac mini

ToolPiper downloads, manages, and runs local models on Apple Silicon. Free, and nothing leaves the machine.

Get ToolPiperCompare against another Mac