---
title: "Mac Studio M3 Ultra: What AI Models Can It Run? | ModelPiper"
description: "Local AI on the M3 Ultra Mac Studio: 819 GB/s memory bandwidth and 96 to 512 GB of unified memory. See which LLMs fit, at which quantization, and how fast they run."
canonical: "https://modelpiper.com/mac-studio/m3-ultra"
---

# Mac Studio M3 Ultra: What AI Models Can It Run? | ModelPiper

> Local AI on the M3 Ultra Mac Studio: 819 GB/s memory bandwidth and 96 to 512 GB of unified memory. See which LLMs fit, at which quantization, and how fast they run.

[← All Mac Studio models](/mac-studio)

# Mac Studio M3 Ultra

The M3 Ultra Mac Studio runs local models at 819 GB/s of memory bandwidth with 96 to 512 GB of unified memory. On Apple Silicon that memory is shared with the GPU, so the whole pool is available for weights: at 512 GB you can hold roughly a 696B dense model at Q4. Two Max dies fused together: the widest memory bus and the highest capacity Apple sells. This is the tier that runs frontier-size open weights locally.

## Specifications

ChipApple M3 Ultra

CPU cores28 or 32

GPU cores60 or 80

Unified memory96, 256, or 512 GB

Memory bandwidth819 GB/s

Neural Engine36 TOPS

Released2025

AvailabilitySold new by Apple

Memory bandwidth is faster than 94% of the Apple Silicon chips shipped in a Mac, against a 819 GB/s peak.

Its memory ceiling is above 94% 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.

GPU cores

60-core 80-core

Unified memory

96 GB

Apple couples memory to the core count on this chip, so the options change with the bin above.

## What each memory option runs

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

[96 GB](#ram-96gb)[256 GB](#ram-256gb)[512 GB](#ram-512gb)

### 96 GB unified memory

84 GB usable for weights · 819 GB/s

3,556 of 3,641 models fit, and 3,503 of them run with headroom rather than as a squeeze.

Largest model at Q4

[Qwen3.5 122B A10B](/fit/qwen-qwen3-5-122b-a10b) · 125.09B

Best all-round pick

[Qwen3.6 35B A3B](/fit/qwen-qwen3-6-35b-a3b) · Q8\_0 · ~143 tok/s

### 256 GB unified memory

225 GB usable for weights · 819 GB/s

3,608 of 3,641 models fit, and 3,581 of them run with headroom rather than as a squeeze.

Largest model at Q4

[DeepSeek V3.2 REAP 345B A37B](/fit/cerebras-deepseek-v3-2-reap-345b-a37b) · 344.89B

Best all-round pick

[Qwen3.6 35B A3B](/fit/qwen-qwen3-6-35b-a3b) · Q8\_0 · ~143 tok/s

### 512 GB unified memory

450 GB usable for weights · 819 GB/s

3,637 of 3,641 models fit, and 3,607 of them run with headroom rather than as a squeeze.

Largest model at Q4

[Kimi K2.6 REAP Solidity](/fit/sombra-x-kimi-k2-6-reap-solidity) · 688.12B

Best all-round pick

[Qwen3.6 35B A3B](/fit/qwen-qwen3-6-35b-a3b) · Q8\_0 · ~143 tok/s

## What a 96 GB M3 Ultra Mac Studio can run

Every model in the database against this exact configuration, at 819 GB/s. Ratings and speeds are the same numbers the model pages show.

All Categories

Show All

All Sizes

Best Fit

Showing 3641 of 3641 models

[Qwen3.6 35B A3B](/fit/qwen-qwen3-6-35b-a3b)

Multimodal · Alibaba · 2026-04-15

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 4B](/fit/qwen-qwen3-5-4b)

Multimodal · Alibaba · 2026-02-27

Q8\_0Excellent

5.7 GB6% of RAM~92 tok/sEstimated4.66B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 12b it](/fit/google-gemma-4-12b-it)

Multimodal · google · 2026-05

Q8\_0Excellent

13.8 GB14% of RAM~36 tok/sEstimated11.96B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 0.8B](/fit/qwen-qwen3-5-0-8b)

Multimodal · Alibaba · 2026-02-28

Q8\_0Excellent

1.5 GB2% of RAM~493 tok/sEstimated0.87B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 35B A3B](/fit/qwen-qwen3-5-35b-a3b)

Multimodal · Alibaba · 2026-02-24

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 2B](/fit/qwen-qwen3-5-2b)

Multimodal · Alibaba · 2026-02-28

Q8\_0Excellent

3.0 GB3% of RAM~189 tok/sEstimated2.27B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 0.8B Base](/fit/qwen-qwen3-5-0-8b-base)

Multimodal · Alibaba · 2026-02-28

Q8\_0Excellent

1.5 GB2% of RAM~493 tok/sEstimated0.87B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 4B Base](/fit/qwen-qwen3-5-4b-base)

Multimodal · Alibaba · 2026-02-27

Q8\_0Excellent

5.7 GB6% of RAM~92 tok/sEstimated4.66B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 2B Base](/fit/qwen-qwen3-5-2b-base)

Multimodal · Alibaba · 2026-02-28

Q8\_0Excellent

3.0 GB3% of RAM~189 tok/sEstimated2.27B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 9B Claude 4.6 Opus Reasoning Distilled v2](/fit/jackrong-qwen3-5-9b-claude-4-6-opus-reasoning-distilled-v2)

Reasoning · jackrong · 2026-03-16

Q8\_0Excellent

11.3 GB12% of RAM~44 tok/sEstimated9.65B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 35B A3B Claude 4.6 Opus Reasoning Distilled](/fit/jackrong-qwen3-5-35b-a3b-claude-4-6-opus-reasoning-distilled)

Reasoning · jackrong · 2026-03-07

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 9B](/fit/qwen-qwen3-5-9b)

Multimodal · Alibaba · 2026-02-27

Q8\_0Excellent

11.3 GB12% of RAM~44 tok/sEstimated9.65B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 9B Base](/fit/qwen-qwen3-5-9b-base)

Multimodal · Alibaba · 2026-02-26

Q8\_0Excellent

11.3 GB12% of RAM~44 tok/sEstimated9.65B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 1.2B](/fit/liquidai-lfm2-1-2b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 ColBERT 350M](/fit/liquidai-lfm2-colbert-350m)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 1.2B Thinking](/fit/liquidai-lfm2-5-1-2b-thinking)

Reasoning · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 8B A1B](/fit/liquidai-lfm2-8b-a1b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

9.8 GB10% of RAMBenchmark needed8.3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 24B A2B](/fit/liquidai-lfm2-24b-a2b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

27.1 GB28% of RAMBenchmark needed23.84B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 1.2B Base](/fit/liquidai-lfm2-5-1-2b-base)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 350M](/fit/liquidai-lfm2-350m)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 2.6B](/fit/liquidai-lfm2-2-6b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

3.4 GB4% of RAM~167 tok/sEstimated2.57B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 700M](/fit/liquidai-lfm2-700m)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.3 GB1% of RAM~580 tok/sEstimated0.74B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 1.2B Extract](/fit/liquidai-lfm2-1-2b-extract)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 350M Extract](/fit/liquidai-lfm2-350m-extract)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 2.6B Transcript](/fit/liquidai-lfm2-2-6b-transcript)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

3.4 GB4% of RAM~167 tok/sEstimated2.57B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 350M PII Extract JP](/fit/liquidai-lfm2-350m-pii-extract-jp)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 1.2B Tool](/fit/liquidai-lfm2-1-2b-tool)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 350M ENJP MT](/fit/liquidai-lfm2-350m-enjp-mt)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 1.2B RAG](/fit/liquidai-lfm2-1-2b-rag)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 350M Math](/fit/liquidai-lfm2-350m-math)

Reasoning · Liquid AI · 2025-11-28

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LaSER Qwen3 8B](/fit/alibaba-nlp-laser-qwen3-8b)

General · alibaba-nlp · 2026-03-31

Q8\_0Excellent

9.6 GB10% of RAM~52 tok/sEstimated8.19B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.0 h small](/fit/ibm-granite-granite-4-0-h-small)

General · ibm-granite · 2025-09-16

Q8\_0Excellent

36.4 GB38% of RAMBenchmark needed32.21B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.0 h tiny](/fit/ibm-granite-granite-4-0-h-tiny)

General · ibm-granite · 2025-09-16

Q8\_0Excellent

8.2 GB9% of RAMBenchmark needed6.94B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 1.2B Instruct](/fit/liquidai-lfm2-5-1-2b-instruct)

Chat · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 2.6B Exp](/fit/liquidai-lfm2-2-6b-exp)

Chat · Liquid AI · 2025-11-28

Q8\_0Excellent

3.4 GB4% of RAM~167 tok/sEstimated2.57B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.0 h micro](/fit/ibm-granite-granite-4-0-h-micro)

General · ibm-granite · 2025-09-16

Q8\_0Excellent

4.1 GB4% of RAM~134 tok/sEstimated3.19B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 1.2B JP](/fit/liquidai-lfm2-5-1-2b-jp)

Chat · Liquid AI · 2025-11-28

Q8\_0Excellent

1.8 GB2% of RAM~367 tok/sEstimated1.17B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VAETKI 7B A1B](/fit/nc-ai-consortium-vaetki-7b-a1b)

General · NCAI · 2025-12-29

Q8\_0Excellent

8.6 GB9% of RAMBenchmark needed7.25B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VAETKI 20B A2B](/fit/nc-ai-consortium-vaetki-20b-a2b)

General · NCAI · 2025-12-29

Q8\_0Excellent

22.4 GB23% of RAMBenchmark needed19.6B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 26B A4B it](/fit/google-gemma-4-26b-a4b-it)

Multimodal · Google · 2025-07-30

Q8\_0Excellent

29.5 GB31% of RAMBenchmark needed26B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 E4B it](/fit/google-gemma-4-e4b-it)

Multimodal · Google · 2025-07-30

Q8\_0Excellent

9.4 GB10% of RAM~54 tok/sEstimated8B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 E2B it](/fit/google-gemma-4-e2b-it)

Multimodal · Google · 2025-07-30

Q8\_0Excellent

6.2 GB6% of RAM~84 tok/sEstimated5.1B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[SmolLM3 3B](/fit/huggingfacetb-smollm3-3b)

Reasoning · HuggingFace · 2025-07-08

Q8\_0Excellent

3.8 GB4% of RAM~143 tok/sEstimated3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 3n E2B it](/fit/google-gemma-3n-e2b-it)

Multimodal · Google · 2025-06-25

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 Audio 1.5B](/fit/liquidai-lfm2-5-audio-1-5b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

2.2 GB2% of RAM~286 tok/sEstimated1.5B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 Audio 1.5B](/fit/liquidai-lfm2-audio-1-5b)

General · Liquid AI · 2025-11-28

Q8\_0Excellent

2.2 GB2% of RAM~286 tok/sEstimated1.5B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VAETKI VL 7B A1B](/fit/nc-ai-consortium-vaetki-vl-7b-a1b)

Multimodal · NCAI · 2025-12-29

Q8\_0Excellent

9.0 GB9% of RAMBenchmark needed7.58B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[NVIDIA Nemotron Nano 9B v2](/fit/nvidia-nvidia-nemotron-nano-9b-v2)

Reasoning · NVIDIA · 2025-06-01

Q8\_0Excellent

10.5 GB11% of RAM~48 tok/sEstimated9B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 VL 450M](/fit/liquidai-lfm2-vl-450m)

Multimodal · Liquid AI · 2025-11-28

Q8\_0Excellent

1.0 GB1% of RAM~953 tok/sEstimated0.45B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 VL 1.6B](/fit/liquidai-lfm2-5-vl-1-6b)

Multimodal · Liquid AI · 2025-11-28

Q8\_0Excellent

2.3 GB2% of RAM~268 tok/sEstimated1.6B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 VL 3B](/fit/liquidai-lfm2-vl-3b)

Multimodal · Liquid AI · 2025-11-28

Q8\_0Excellent

3.8 GB4% of RAM~143 tok/sEstimated3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2 VL 1.6B](/fit/liquidai-lfm2-vl-1-6b)

Multimodal · Liquid AI · 2025-11-28

Q8\_0Excellent

2.3 GB2% of RAM~272 tok/sEstimated1.58B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[EXAONE Deep 2.4B](/fit/lgai-exaone-exaone-deep-2-4b)

General · lgai-exaone · 2025-03-12

Q8\_0Excellent

3.2 GB3% of RAM~178 tok/sEstimated2.41B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[DeepSeek R1 Distill Qwen 7B](/fit/deepseek-ai-deepseek-r1-distill-qwen-7b)

Reasoning · DeepSeek · 2025-01-20

Q8\_0Excellent

9.0 GB9% of RAM~56 tok/sEstimated7.62B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled](/fit/jackrong-qwen3-5-27b-claude-4-6-opus-reasoning-distilled)

Reasoning · jackrong · 2026-02-27

Q8\_0Excellent

31.5 GB33% of RAM~15 tok/sEstimated27.78B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[EXAONE 4.0 1.2B](/fit/lgai-exaone-exaone-4-0-1-2b)

General · LG AI · 2025-07-15

Q8\_0Excellent

1.8 GB2% of RAM~358 tok/sEstimated1.2B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Jolia](/fit/raidium-jolia)

General · raidium · 2026-06-15

Q8\_0Excellent

0.5 GB1% of RAM~21,450 tok/sEstimated0.02B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[embeddinggemma GTAIDE 300m 2605](/fit/taide-embeddinggemma-gtaide-300m-2605)

Embedding · taide · 2026-06-12

Q8\_0Excellent

0.8 GB1% of RAM~1,430 tok/sEstimated0.3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 0.6B](/fit/qwen-qwen3-0-6b)

General · Alibaba · 2025-04-27

Q8\_0Excellent

1.3 GB1% of RAM~572 tok/sEstimated0.75B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 4B](/fit/qwen-qwen3-4b)

General · Alibaba · 2025-04-27

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4.02B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 8B](/fit/qwen-qwen3-8b)

General · Alibaba · 2025-04-27

Q8\_0Excellent

9.6 GB10% of RAM~52 tok/sEstimated8.19B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gpt oss 20b](/fit/openai-gpt-oss-20b)

General · openai

Q8\_0Excellent

24.5 GB26% of RAMBenchmark needed21.51B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.6 27B](/fit/qwen-qwen3-6-27b)

Multimodal · Alibaba · 2026-04-21

Q8\_0Excellent

31.5 GB33% of RAM~15 tok/sEstimated27.78B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 1.7B](/fit/qwen-qwen3-1-7b)

General · Alibaba · 2025-04-27

Q8\_0Excellent

2.8 GB3% of RAM~211 tok/sEstimated2.03B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 VL 4B Instruct](/fit/qwen-qwen3-vl-4b-instruct)

Multimodal · Alibaba

Q8\_0Excellent

5.5 GB6% of RAM~97 tok/sEstimated4.44B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[GLM OCR](/fit/zai-org-glm-ocr)

Multimodal · zai-org

Q8\_0Excellent

2.0 GB2% of RAM~323 tok/sEstimated1.33B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 30B A3B](/fit/qwen-qwen3-30b-a3b)

General · Alibaba · 2025-04-27

Q8\_0Excellent

34.6 GB36% of RAMBenchmark needed30.53B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 3 12b it](/fit/google-gemma-3-12b-it)

Multimodal · Google · 2025-03-01

Q8\_0Excellent

13.9 GB14% of RAM~36 tok/sEstimated12B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 VL 2B Instruct](/fit/qwen-qwen3-vl-2b-instruct)

Multimodal · Alibaba

Q8\_0Excellent

2.9 GB3% of RAM~201 tok/sEstimated2.13B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 Coder 30B A3B Instruct](/fit/qwen-qwen3-coder-30b-a3b-instruct)

Coding · Alibaba

Q8\_0Excellent

34.6 GB36% of RAMBenchmark needed30.53B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[GLM 4.7 Flash](/fit/zai-org-glm-4-7-flash)

General · zai-org

Q8\_0Excellent

35.3 GB37% of RAMBenchmark needed31.22B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[chandra ocr 2](/fit/datalab-to-chandra-ocr-2)

Multimodal · datalab-to

Q8\_0Excellent

6.4 GB7% of RAM~81 tok/sEstimated5.3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Rio 3.0 Open Mini](/fit/prefeitura-rio-rio-3-0-open-mini)

General · prefeitura-rio

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4.02B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Phi 3.5 vision instruct](/fit/microsoft-phi-3-5-vision-instruct)

Multimodal · Microsoft

Q8\_0Excellent

5.1 GB5% of RAM~103 tok/sEstimated4.15B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 VL 4B Thinking](/fit/qwen-qwen3-vl-4b-thinking)

General · Alibaba

Q8\_0Excellent

5.5 GB6% of RAM~97 tok/sEstimated4.44B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen2.5 1.5B](/fit/qwen-qwen2-5-1-5b)

General · Alibaba

Q8\_0Excellent

2.2 GB2% of RAM~279 tok/sEstimated1.54B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen2 0.5B](/fit/qwen-qwen2-0-5b)

General · Alibaba

Q8\_0Excellent

1.0 GB1% of RAM~876 tok/sEstimated0.49B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[diffusiongemma 26B A4B it](/fit/google-diffusiongemma-26b-a4b-it)

Multimodal · Google

Q8\_0Excellent

29.3 GB31% of RAMBenchmark needed25.82B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[MiniCPM V 4.6](/fit/openbmb-minicpm-v-4-6)

Multimodal · openbmb

Q8\_0Excellent

2.0 GB2% of RAM~330 tok/sEstimated1.3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 4B Thinking 2507](/fit/qwen-qwen3-4b-thinking-2507)

General · Alibaba

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4.02B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[DeepSeek R1 Distill Qwen 1.5B](/fit/deepseek-ai-deepseek-r1-distill-qwen-1-5b)

Reasoning · DeepSeek

Q8\_0Excellent

2.5 GB3% of RAM~241 tok/sEstimated1.78B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[dots.mocr](/fit/rednote-hilab-dots-mocr)

Multimodal · rednote-hilab

Q8\_0Excellent

3.9 GB4% of RAM~141 tok/sEstimated3.04B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 VL 30B A3B Instruct](/fit/qwen-qwen3-vl-30b-a3b-instruct)

Multimodal · Alibaba

Q8\_0Excellent

35.2 GB37% of RAMBenchmark needed31.07B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen2.5 7B](/fit/qwen-qwen2-5-7b)

General · Alibaba

Q8\_0Excellent

9.0 GB9% of RAM~56 tok/sEstimated7.62B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Phi 4 multimodal instruct](/fit/microsoft-phi-4-multimodal-instruct)

Multimodal · Microsoft · 2025-04-01

Q8\_0Excellent

16.1 GB17% of RAM~31 tok/sEstimated14B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[DeepSeek R1 0528 Qwen3 8B](/fit/deepseek-ai-deepseek-r1-0528-qwen3-8b)

Reasoning · DeepSeek

Q8\_0Excellent

9.6 GB10% of RAM~52 tok/sEstimated8.19B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[UI TARS 1.5 7B](/fit/bytedance-seed-ui-tars-1-5-7b)

Multimodal · bytedance-seed

Q8\_0Excellent

9.7 GB10% of RAM~52 tok/sEstimated8.29B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Idefics3 8B Llama3](/fit/huggingfacem4-idefics3-8b-llama3)

Multimodal · huggingfacem4

Q8\_0Excellent

9.9 GB10% of RAM~51 tok/sEstimated8.46B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VLM2Vec Full](/fit/tiger-lab-vlm2vec-full)

General · tiger-lab

Q8\_0Excellent

5.1 GB5% of RAM~103 tok/sEstimated4.15B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[DeepSeek V2 Lite](/fit/deepseek-ai-deepseek-v2-lite)

General · DeepSeek

Q8\_0Excellent

18.0 GB19% of RAMBenchmark needed15.71B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 26B A4B](/fit/google-gemma-4-26b-a4b)

Multimodal · Google

Q8\_0Excellent

30.1 GB31% of RAMBenchmark needed26.54B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[surya ocr 2](/fit/datalab-to-surya-ocr-2)

Multimodal · datalab-to

Q8\_0Excellent

1.3 GB1% of RAM~622 tok/sEstimated0.69B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.1 3b](/fit/ibm-granite-granite-4-1-3b)

General · ibm-granite

Q8\_0Excellent

4.3 GB4% of RAM~126 tok/sEstimated3.4B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[DeepSeek R1 Distill Llama 8B](/fit/deepseek-ai-deepseek-r1-distill-llama-8b)

Reasoning · DeepSeek

Q8\_0Excellent

9.5 GB10% of RAM~53 tok/sEstimated8.03B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[OTel LLM 4B IT](/fit/farbodtavakkoli-otel-llm-4b-it)

General · farbodtavakkoli

Q8\_0Excellent

5.3 GB6% of RAM~100 tok/sEstimated4.3B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Kimi VL A3B Instruct](/fit/moonshotai-kimi-vl-a3b-instruct)

Multimodal · moonshotai

Q8\_0Excellent

18.8 GB20% of RAMBenchmark needed16.41B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[MiniCPM5 1B](/fit/openbmb-minicpm5-1b)

General · openbmb

Q8\_0Excellent

1.7 GB2% of RAM~397 tok/sEstimated1.08B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[OTel LLM 8.3B IT](/fit/farbodtavakkoli-otel-llm-8-3b-it)

General · farbodtavakkoli

Q8\_0Excellent

8.3 GB9% of RAM~62 tok/sEstimated6.97B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[dots.ocr](/fit/rednote-hilab-dots-ocr)

Multimodal · rednote-hilab

Q8\_0Excellent

3.9 GB4% of RAM~141 tok/sEstimated3.04B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Ilama 3.2 1B](/fit/hmellor-ilama-3-2-1b)

General · hmellor

Q8\_0Excellent

1.9 GB2% of RAM~346 tok/sEstimated1.24B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[distil lfm25 shellper](/fit/distil-labs-distil-lfm25-shellper)

General · distil-labs

Q8\_0Excellent

0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Nanonets OCR2 3B](/fit/nanonets-nanonets-ocr2-3b)

Multimodal · nanonets

Q8\_0Excellent

4.7 GB5% of RAM~114 tok/sEstimated3.75B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 26B A4B it uncensored](/fit/trevorjs-gemma-4-26b-a4b-it-uncensored)

General · trevorjs

Q8\_0Excellent

29.3 GB31% of RAMBenchmark needed25.81B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[OTel LLM 1.2B IT](/fit/farbodtavakkoli-otel-llm-1-2b-it)

General · farbodtavakkoli

Q8\_0Excellent

1.8 GB2% of RAM~355 tok/sEstimated1.21B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Laguna XS.2](/fit/poolside-laguna-xs-2)

General · poolside

Q8\_0Excellent

37.8 GB39% of RAMBenchmark needed33.44B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Karnak 40B v1.0](/fit/applied-innovation-center-karnak-40b-v1-0)

General · applied-innovation-center

Q8\_0Excellent

45.9 GB48% of RAMBenchmark needed40.67B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[RolmOCR](/fit/reducto-rolmocr)

Multimodal · reducto

Q8\_0Excellent

9.7 GB10% of RAM~52 tok/sEstimated8.29B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 30B A3B Thinking 2507](/fit/qwen-qwen3-30b-a3b-thinking-2507)

General · Alibaba

Q8\_0Excellent

34.6 GB36% of RAMBenchmark needed30.53B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[EXAONE Deep 7.8B](/fit/lgai-exaone-exaone-deep-7-8b)

General · lgai-exaone

Q8\_0Excellent

9.2 GB10% of RAM~55 tok/sEstimated7.82B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LFM2.5 8B A1B](/fit/liquidai-lfm2-5-8b-a1b)

General · Liquid AI

Q8\_0Excellent

9.9 GB10% of RAMBenchmark needed8.47B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Nanonets OCR s](/fit/nanonets-nanonets-ocr-s)

General · nanonets

Q8\_0Excellent

4.7 GB5% of RAM~114 tok/sEstimated3.75B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Nanbeige4.1 3B](/fit/nanbeige-nanbeige4-1-3b)

General · nanbeige

Q8\_0Excellent

4.9 GB5% of RAM~109 tok/sEstimated3.93B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.6 35B A3B Claude 4.7 Opus Reasoning Distilled](/fit/lordx64-qwen3-6-35b-a3b-claude-4-7-opus-reasoning-distilled)

Reasoning · lordx64

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[LLaVA OneVision 1.5 4B Base](/fit/lmms-lab-llava-onevision-1-5-4b-base)

General · lmms-lab

Q8\_0Excellent

5.8 GB6% of RAM~91 tok/sEstimated4.74B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite vision 4.1 4b](/fit/ibm-granite-granite-vision-4-1-4b)

Multimodal · ibm-granite

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Molmo2 O 7B](/fit/allenai-molmo2-o-7b)

Multimodal · allenai

Q8\_0Excellent

9.2 GB10% of RAM~55 tok/sEstimated7.76B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[OTel LLM 3B IT](/fit/farbodtavakkoli-otel-llm-3b-it)

General · farbodtavakkoli

Q8\_0Excellent

5.2 GB5% of RAM~101 tok/sEstimated4.25B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.0 3b vision](/fit/ibm-granite-granite-4-0-3b-vision)

General · ibm-granite

Q8\_0Excellent

5.0 GB5% of RAM~107 tok/sEstimated4B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gpt oss 20b Derestricted](/fit/arliai-gpt-oss-20b-derestricted)

General · arliai

Q8\_0Excellent

23.8 GB25% of RAMBenchmark needed20.91B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.0 tiny preview](/fit/ibm-granite-granite-4-0-tiny-preview)

General · ibm-granite

Q8\_0Excellent

7.9 GB8% of RAMBenchmark needed6.67B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[reformer crime and punishment](/fit/google-reformer-crime-and-punishment)

General · Google

Q8\_0Excellent

0.5 GB1% of RAM~42,900 tok/sEstimated0.01B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Kimi VL A3B Thinking](/fit/moonshotai-kimi-vl-a3b-thinking)

Multimodal · moonshotai

Q8\_0Excellent

18.8 GB20% of RAMBenchmark needed16.41B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VideoLLaMA3 2B Image HF](/fit/lkhl-videollama3-2b-image-hf)

Multimodal · lkhl

Q8\_0Excellent

2.7 GB3% of RAM~219 tok/sEstimated1.96B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[typhoon ocr 3b](/fit/typhoon-ai-typhoon-ocr-3b)

Multimodal · typhoon-ai

Q8\_0Excellent

4.7 GB5% of RAM~114 tok/sEstimated3.75B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[ReaderLM v2](/fit/jinaai-readerlm-v2)

General · jinaai

Q8\_0Excellent

2.2 GB2% of RAM~279 tok/sEstimated1.54B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Huihui Qwen3.5 35B A3B Claude 4.6 Opus abliterated](/fit/huihui-ai-huihui-qwen3-5-35b-a3b-claude-4-6-opus-abliterated)

General · huihui-ai

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[sarvam 30b uncensored](/fit/aoxo-sarvam-30b-uncensored)

General · aoxo

Q8\_0Excellent

36.4 GB38% of RAMBenchmark needed32.15B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3 VL 2B Thinking](/fit/qwen-qwen3-vl-2b-thinking)

General · Alibaba

Q8\_0Excellent

2.9 GB3% of RAM~201 tok/sEstimated2.13B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Josiefied Qwen3 VL 4B Instruct abliterated beta v1](/fit/goekdeniz-guelmez-josiefied-qwen3-vl-4b-instruct-abliterated-beta-v1)

Multimodal · goekdeniz-guelmez

Q8\_0Excellent

5.5 GB6% of RAM~97 tok/sEstimated4.44B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gemma 4 26B A4B it heretic](/fit/coder3101-gemma-4-26b-a4b-it-heretic)

Coding · coder3101

Q8\_0Excellent

29.3 GB31% of RAMBenchmark needed25.81B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[dots.ocr 1.5](/fit/kristaller486-dots-ocr-1-5)

General · kristaller486

Q8\_0Excellent

3.9 GB4% of RAM~141 tok/sEstimated3.04B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 35B A3B uncensored heretic](/fit/llmfan46-qwen3-5-35b-a3b-uncensored-heretic)

General · llmfan46

Q8\_0Excellent

39.7 GB41% of RAMBenchmark needed35.11B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[typhoon ocr1.5 2b](/fit/typhoon-ai-typhoon-ocr1-5-2b)

Reasoning · typhoon-ai

Q8\_0Excellent

2.9 GB3% of RAM~201 tok/sEstimated2.13B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Holo3 35B A3B](/fit/hcompany-holo3-35b-a3b)

General · hcompany

Q8\_0Excellent

39.7 GB41% of RAMBenchmark needed35.11B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[sarvam 30b](/fit/sarvamai-sarvam-30b)

General · sarvamai

Q8\_0Excellent

36.4 GB38% of RAMBenchmark needed32.15B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Solar Open 100B](/fit/upstage-solar-open-100b)

General · Upstage

Q8\_0Excellent

9.5 GB10% of RAMBenchmark needed8.05B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Yi 6B 200K](/fit/01-ai-yi-6b-200k)

General · 01.ai

Q8\_0Excellent

7.3 GB8% of RAM~71 tok/sEstimated6.06B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[HTML Pruner Phi 3.8B](/fit/zstanjj-html-pruner-phi-3-8b)

General · zstanjj

Q8\_0Excellent

4.8 GB5% of RAM~112 tok/sEstimated3.82B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 4.1 3b base](/fit/ibm-granite-granite-4-1-3b-base)

General · ibm-granite

Q8\_0Excellent

4.3 GB4% of RAM~126 tok/sEstimated3.4B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Nemotron Labs Diffusion 3B](/fit/nvidia-nemotron-labs-diffusion-3b)

General · nvidia

Q8\_0Excellent

4.8 GB5% of RAM~112 tok/sEstimated3.83B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.6 35B A3B uncensored heretic](/fit/llmfan46-qwen3-6-35b-a3b-uncensored-heretic)

General · llmfan46

Q8\_0Excellent

39.7 GB41% of RAMBenchmark needed35.11B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[VibeThinker 3B](/fit/weiboai-vibethinker-3b)

General · weiboai

Q8\_0Excellent

3.9 GB4% of RAM~139 tok/sEstimated3.09B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Nemotron Cascade 2 30B A3B](/fit/nvidia-nemotron-cascade-2-30b-a3b)

General · nvidia

Q8\_0Excellent

35.7 GB37% of RAMBenchmark needed31.58B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[granite 8b code instruct 128k](/fit/ibm-granite-granite-8b-code-instruct-128k)

Coding · ibm-granite

Q8\_0Excellent

9.5 GB10% of RAM~53 tok/sEstimated8.05B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen3.5 35B A3B Base](/fit/qwen-qwen3-5-35b-a3b-base)

General · Alibaba

Q8\_0Excellent

40.6 GB42% of RAMBenchmark needed35.95B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Qwen2 7B](/fit/qwen-qwen2-7b)

General · Alibaba

Q8\_0Excellent

9.0 GB9% of RAM~56 tok/sEstimated7.62B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[libra v1.0 3b](/fit/x-izhang-libra-v1-0-3b)

General · x-izhang

Q8\_0Excellent

4.1 GB4% of RAM~132 tok/sEstimated3.25B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[Rex Omni](/fit/idea-research-rex-omni)

General · idea-research

Q8\_0Excellent

5.0 GB5% of RAM~105 tok/sEstimated4.07B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[gpt oss safeguard 120b](/fit/openai-gpt-oss-safeguard-120b)

General · openai

Q8\_0Excellent

3.1 GB3% of RAMBenchmark needed2.37B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

[PaddleOCR VL 1.5](/fit/paddlepaddle-paddleocr-vl-1-5)

General · paddlepaddle

Q8\_0Excellent

1.6 GB2% of RAM~447 tok/sEstimated0.96B params

What Mac for this model? [Run with ToolPiper](/toolpiper)

Show more (3491 remaining)

## 60-core vs 80-core GPU

Both bins run the same 819 GB/s memory bus, so token generation is the same on either one. The extra cores show up in image and video work, not in tokens per second.

Configuration

Memory bandwidth

Memory options

Models that fit

28-core CPU, 60-core GPU

819 GB/s

96 GB

Identical

32-core CPU, 80-core GPU

819 GB/s

96, 256, 512 GB

Identical

## Measured on the M3 Ultra

Nobody has submitted a benchmark on the M3 Ultra 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](/fit/benchmarks) 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.

[Used market alternative

Mac Studio M2 Ultra

2% less memory bandwidth, 192 GB ceiling instead of 512 GB

](/mac-studio/m2-ultra)

## 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 M3 Ultra's 819 GB/s bus is shared by CPU, GPU, and Neural Engine, so a 512 GB machine can hand almost all of that to a model with no copy across a bus.

The Studio exists for this workload. It carries the widest memory buses and the highest capacities Apple sells, and it runs at full clocks indefinitely. Buy the memory, not the cores: every extra GB raises what you can load, while the core count only moves throughput on models that already fit.

## Common questions

### Can the M3 Ultra Mac Studio run a 70B model?

Yes, at 512 GB. A 70B model at Q4\_K\_M needs about 46 GB including an 8K context, and 512 GB of unified memory leaves about 450 GB for weights once macOS takes its share. At 96 GB it does not fit at any quantization worth running.

### How much unified memory should I get with the M3 Ultra Mac Studio?

Memory is the only spec that changes what you can run at all. 96 GB holds about a 129B model at Q4; 512 GB holds about 696B. 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 M3 Ultra?

Token generation is bandwidth-bound, so M3 Ultra throughput scales with its 819 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 M3 Ultra lands in the tens of tokens per second and a 70B model lands in the single digits.

### Is the 80-core GPU worth it over the 60-core on the M3 Ultra?

Not for LLMs. Both bins run the same 819 GB/s memory bus and take the same memory options, and token generation is bound by bandwidth rather than GPU cores. The extra cores show up in image generation and video work, not in tokens per second.

### Should I buy the M3 Ultra Mac Studio now or wait for the next one?

Buy on the memory you need today. Apple raises memory ceilings slowly and bandwidth in steps, and the M3 Ultra already holds about a 696B model at Q4. If your target model fits in 512 GB, waiting buys throughput rather than capability.

## Run these models on your Mac Studio

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

[Get ToolPiper](/toolpiper)[Compare against another Mac](/fit?chip=m3-ultra&ram=96)

[![ModelPiper](modelpiper-logo-sm.png)ModelPiper](/)

Local AI. Private by default.  
Talk to your Mac.

#### Products

-   [ToolPiper](/toolpiper)
-   [VisionPiper](/visionpiper)
-   [AudioPiper](/audiopiper)
-   [MediaPiper](/docs/mediapiper)
-   [PiperTest](/pipertest)

#### Mac Hardware

-   [MacBook Air](/macbook-air)
-   [MacBook Pro 14"](/macbook-pro-14)
-   [MacBook Pro 16"](/macbook-pro-16)
-   [Mac mini](/mac-mini)
-   [Mac Studio](/mac-studio)
-   [iMac](/imac)

#### Resources

-   [Model Fit](/fit)
-   [Benchmarks](/fit/benchmarks)
-   [Blog](/blog)
-   [Workflows](/workflow)
-   [Docs](/docs)
-   [MCP Tools](/mcp-tools)

#### Company

-   [About](/about)
-   [Press](/press)
-   [Changelog](/changelog)
-   [Pricing](/pricing)
-   [Download](/download)
-   [Privacy](/privacy)
-   [Terms](/terms)

© 2026 ModelPiper. All rights reserved.

My Connections

### AI Providers

No providers yet. Click + to add one.

Services

ToolPiper

VisionPiper

AudioPiper