AI Index

Open-weight models that run on Apple M3 Ultra · 96 GB

  • Unified memory 96 GB
  • Bandwidth 819 GB/s
  • Usable 72.0 GB

Specs: Memory · Mac Studio (2025) - Tech Specs - Apple Support · Bandwidth · Mac Studio (2025) - Tech Specs - Apple Support

FitModelParameters / activePrecisionMemoryEst. speedLongest contextRuns as
Runs wellMing-3.0-audio-flash Ant Group128.2B / 6.5BQ4_070.9 / 72.0 GB101–152 tok/s32KUnified memory
Runs wellLing-3.0-flash Ant Group127.5B / 5.1BQ4_070.5 / 72.0 GB125–188 tok/s40KUnified memory
Runs wellQwen3.5-122B-A10B Alibaba Qwen125.1B / 10BQ4_069.0 / 72.0 GB70–105 tok/s136KUnified memory
Runs wellLing-3.0-flash-VL Ant Group124.8B / 5.5BQ4_069.0 / 72.0 GB117–176 tok/s72KUnified memory
Runs wellgpt-oss-120b OpenAI116.8B / 5.1BMXFP4 (native)maker’s file64.1 / 72.0 GB130–195 tok/s128KUnified memory
Runs wellGLM-4.5-Air Zhipu AI110.5BQ4_062.2 / 72.0 GB37–55 tok/s63KUnified memory
Runs wellGLM-4.5V Zhipu AI107.7B / 12BQ4_060.7 / 72.0 GB49–74 tok/s63KUnified memory
Runs wellGLM-4.6V Zhipu AI107.7BQ4_060.7 / 72.0 GB43–64 tok/s71KUnified memory
Runs wellLing-2.6-flash Ant Group107.5B / 7.4BQ4_K_M64.1 / 72.0 GB85–128 tok/s128KUnified memory
Runs wellQwen3-Next-80B-A3B-Instruct Alibaba Qwen81.3B / 3BQ6_Kmaker’s file64.3 / 72.0 GB156–235 tok/s256KUnified memory
Runs wellQwen3-Next-80B-A3B-Thinking Alibaba Qwen81.3B / 3BQ6_Kmaker’s file64.3 / 72.0 GB156–235 tok/s256KUnified memory
Runs wellQwen3-Coder-Next Alibaba Qwen79.7B / 3BQ6_Kmaker’s file64.3 / 72.0 GB153–230 tok/s256KUnified memory
Runs wellQwen3.5-35B-A3B Alibaba Qwen36B / 3BQ8_037.5 / 72.0 GB122–183 tok/s256KUnified memory
Runs wellQwen3.6-35B-A3B Alibaba Qwen36B / 3BQ8_037.5 / 72.0 GB122–183 tok/s256KUnified memory
Runs wellQwen2.5-VL-32B-Instruct Alibaba Qwen33.5BQ6_K28.9 / 72.0 GB14–21 tok/s125KUnified memory
Runs wellQwen3-VL-32B-Instruct Alibaba Qwen33.4BQ6_K28.8 / 72.0 GB14–21 tok/s128KUnified memory
Runs wellQwen3-VL-32B-Thinking Alibaba Qwen33.4BQ6_K28.8 / 72.0 GB14–21 tok/s128KUnified memory
Runs wellQwen2.5-32B-Instruct Alibaba Qwen32.8BQ6_Kmaker’s file28.3 / 72.0 GB14–21 tok/s32KUnified memory
Runs wellQwen2.5-Coder-32B-Instruct Alibaba Qwen32.8BINT4 (AWQ / GPTQ)maker’s file20.9 / 72.0 GB19–29 tok/s32KUnified memory
Runs wellQwQ-32B Alibaba Qwen32.8BQ6_Kmaker’s file28.3 / 72.0 GB14–21 tok/s40KUnified memory
Runs wellDeepSeek-R1-Distill-Qwen-32B DeepSeek32.8BQ6_K28.3 / 72.0 GB14–21 tok/s32KUnified memory
Runs wellQwen3-32B Alibaba Qwen32.8BQ6_Kmaker’s file28.3 / 72.0 GB14–21 tok/s128KUnified memory
Runs wellQwen3-Omni-30B-A3B-Captioner Alibaba Qwen31.7BBF16 (native)maker’s file62.8 / 72.0 GB41–62 tok/s64KUnified memory
Runs wellGemma 4 31B IT Google31.3BQ6_K27.1 / 72.0 GB15–22 tok/s250KUnified memory
Runs wellGLM-4.7-Flash Zhipu AI31.2BBF16 (native)maker’s file61.5 / 72.0 GB38–57 tok/s195KUnified memory
Runs wellQwen3-VL-30B-A3B-Instruct Alibaba Qwen31.1BBF16 (native)maker’s file61.5 / 72.0 GB48–72 tok/s120KUnified memory
Runs wellQwen3-VL-30B-A3B-Thinking Alibaba Qwen31.1BBF16 (native)maker’s file61.5 / 72.0 GB48–72 tok/s120KUnified memory
Runs wellQwen3-30B-A3B-Instruct-2507 Alibaba Qwen30.5B / 3.3BBF16 (native)maker’s file60.5 / 72.0 GB55–83 tok/s256KUnified memory
Runs wellQwen3-30B-A3B-Thinking-2507 Alibaba Qwen30.5B / 3.3BBF16 (native)maker’s file60.5 / 72.0 GB55–83 tok/s256KUnified memory
Runs wellQwen3-30B-A3B Alibaba Qwen30.5B / 3.3BBF16 (native)maker’s file60.5 / 72.0 GB55–83 tok/s128KUnified memory
Runs wellQwen3-Coder-30B-A3B-Instruct Alibaba Qwen30.5B / 3.3BBF16 (native)maker’s file60.5 / 72.0 GB55–83 tok/s131KUnified memory
Runs wellHy-Memory-30B-A3B Tencent30.1B / 3BBF16 (native)maker’s file59.6 / 72.0 GB60–90 tok/s63KUnified memory
Runs wellGranite Switch 4.1 30B IBM30BFP831.4 / 72.0 GB13–19 tok/s128KUnified memory
Runs wellQwen3.5-27B Alibaba Qwen27.8BQ8_029.4 / 72.0 GB14–20 tok/s256KUnified memory
Runs wellQwen3.6-27B Alibaba Qwen27.8BQ8_029.4 / 72.0 GB14–20 tok/s256KUnified memory
Runs wellQwen3.8-27B Alibaba Qwen27.8BQ8_029.4 / 72.0 GB14–20 tok/s256KUnified memory
Runs wellGemma 4 26B A4B IT Google25.8B / 3.8BBF16 (native)maker’s file51.0 / 72.0 GB50–75 tok/s250KUnified memory
Runs wellMagistral Small 1.2 Mistral AI24BQ8_0maker’s file25.8 / 72.0 GB16–23 tok/s125KUnified memory
Runs wellDevstral Small 1.0 Mistral AI23.6BQ8_025.7 / 72.0 GB16–23 tok/s125KUnified memory
Runs wellDevstral Small 1.1 Mistral AI23.6BQ8_025.7 / 72.0 GB16–23 tok/s125KUnified memory
Runs wellMagistral Small 1.0 Mistral AI23.6BQ8_025.7 / 72.0 GB16–23 tok/s39KUnified memory
Runs wellMagistral Small 1.1 Mistral AI23.6BQ8_0maker’s file25.8 / 72.0 GB16–23 tok/s39KUnified memory
Runs wellERNIE-4.5-21B-A3B-Thinking Baidu21.8B / 3BBF16 (native)maker’s file43.1 / 72.0 GB63–95 tok/s128KUnified memory
Runs wellGPT OSS 20B OpenAI20.9B / 3.6BMXFP4 (native)maker’s file13.6 / 72.0 GB159–239 tok/s128KUnified memory
Runs wellQwen2.5-14B-Instruct-1M Alibaba Qwen14.8BBF16 (native)maker’s file30.4 / 72.0 GB13–20 tok/s230KUnified memory
Runs wellQwen2.5-14B-Instruct Alibaba Qwen14.8BBF16 (native)maker’s file30.4 / 72.0 GB13–20 tok/s32KUnified memory
Runs wellDeepSeek-R1-Distill-Qwen-14B DeepSeek14.8BBF16 (native)maker’s file30.4 / 72.0 GB13–20 tok/s32KUnified memory
Runs wellQwen3-14B Alibaba Qwen14.8BBF16 (native)maker’s file30.1 / 72.0 GB13–20 tok/s128KUnified memory
Runs wellGemma 3 12B IT Google12.2BBF16 (native)maker’s file26.6 / 72.0 GB15–22 tok/s128KUnified memory
Runs wellQwen2.5-Omni-7B Alibaba Qwen10.7BBF16 (native)maker’s file22.3 / 72.0 GB18–27 tok/s32KUnified memory
Runs wellGLM-4.6V-Flash Zhipu AI10.3BBF16 (native)maker’s file20.4 / 72.0 GB20–29 tok/s125KUnified memory
Runs wellQwen3.5-9B Alibaba Qwen9.7BBF16 (native)maker’s file19.1 / 72.0 GB21–31 tok/s256KUnified memory
Runs wellQianfan-VL-8B Baidu8.8BBF16 (native)maker’s file18.2 / 72.0 GB22–33 tok/s32KUnified memory
Runs wellGranite Switch 4.1 8B IBM8.8BBF16 (native)maker’s file18.5 / 72.0 GB22–32 tok/s128KUnified memory
Runs wellQwen3-VL-8B-Instruct Alibaba Qwen8.8BBF16 (native)maker’s file18.3 / 72.0 GB22–33 tok/s128KUnified memory
Runs wellQwen3-VL-8B-Thinking Alibaba Qwen8.8BBF16 (native)maker’s file18.3 / 72.0 GB22–33 tok/s128KUnified memory
Runs wellpplx-embed-v2-context-9b Perplexity8.4BQ8_09.1 / 72.0 GB45–67 tok/s256KUnified memory
Runs wellQwen2.5-VL-7B-Instruct Alibaba Qwen8.3BBF16 (native)maker’s file16.7 / 72.0 GB24–36 tok/s125KUnified memory
Runs wellQwen3-8B Alibaba Qwen8.2BBF16 (native)maker’s file17.1 / 72.0 GB23–35 tok/s128KUnified memory
Runs wellDeepSeek-R1-Distill-Llama-8B DeepSeek8BBF16 (native)maker’s file16.7 / 72.0 GB24–36 tok/s125KUnified memory
Runs wellGemma 4 E4B Google8BBF16 (native)maker’s file15.8 / 72.0 GB25–38 tok/s125KUnified memory
Runs wellLing-3.0-tiny Ant Group7.9B / 1.3BBF16 (native)maker’s file15.6 / 72.0 GB145–217 tok/s128KUnified memory
Runs wellQwen2.5-7B-Instruct-1M Alibaba Qwen7.6BBF16 (native)maker’s file15.3 / 72.0 GB26–39 tok/s977KUnified memory
Runs wellQwen2.5-7B-Instruct Alibaba Qwen7.6BBF16 (native)maker’s file15.3 / 72.0 GB26–39 tok/s32KUnified memory
Runs wellDeepSeek-R1-Distill-Qwen-7B DeepSeek7.6BBF16 (native)maker’s file15.3 / 72.0 GB26–39 tok/s32KUnified memory
Runs wellMistral-7B-Instruct-v0.1 Mistral AI7.2BBF16 (native)maker’s file15.2 / 72.0 GB26–39 tok/s32KUnified memory
Runs wellGemma 4 E2B Google5.1BBF16 (native)maker’s file10.1 / 72.0 GB40–60 tok/s125KUnified memory
Runs wellWeVisOne-VL-4B Tencent4.8BBF16 (native)maker’s file10.6 / 72.0 GB38–57 tok/s256KUnified memory
Runs wellQianfan-OCR Baidu4.7BBF16 (native)maker’s file10.5 / 72.0 GB38–57 tok/s31KUnified memory
Runs wellQwen3.5-4B Alibaba Qwen4.7BBF16 (native)maker’s file9.4 / 72.0 GB43–64 tok/s256KUnified memory
Runs wellQwen3-VL-4B-Instruct Alibaba Qwen4.4BBF16 (native)maker’s file9.9 / 72.0 GB41–61 tok/s250KUnified memory
Runs wellQwen3-4B-Instruct-2507 Alibaba Qwen4BBF16 (native)maker’s file9.1 / 72.0 GB44–66 tok/s256KUnified memory
Runs wellQwen3-4B Alibaba Qwen4BBF16 (native)maker’s file9.1 / 72.0 GB44–66 tok/s128KUnified memory
Runs wellDeepSeek-OCR DeepSeek3.3BBF16 (native)maker’s file7.2 / 72.0 GB147–221 tok/s8KUnified memory
Runs wellGranite-4.0-H-Micro IBM3.2BBF16 (native)maker’s file6.5 / 72.0 GB63–95 tok/s128KUnified memory
Runs wellQwen3-1.7B Alibaba Qwen2BBF16 (native)maker’s file5.2 / 72.0 GB82–123 tok/s40KUnified memory
Runs wellDeepSeek-R1-Distill-Qwen-1.5B DeepSeek1.8BBF16 (native)maker’s file4.0 / 72.0 GB108–162 tok/s125KUnified memory
Runs wellGLM-OCR Zhipu AI1.3BBF16 (native)maker’s file3.5 / 72.0 GB128–193 tok/s128KUnified memory
Runs wellQwen3-0.6B Alibaba Qwen0.8BBF16 (native)maker’s file2.8 / 72.0 GB168–251 tok/s40KUnified memory
Runs wellEmbeddingGemma 2 Google0.7BBF16 (native)maker’s file2.0 / 72.0 GB260–390 tok/s256KUnified memory
Runs wellQwen3-Embedding-0.6B Alibaba Qwen0.6BBF16 (native)maker’s file2.5 / 72.0 GB192–288 tok/s32KUnified memory
Runs wellbge-reranker-v2-m3 BAAI0.6BFP32 (native)maker’s file3.4 / 72.0 GB133–200 tok/s8KUnified memory
Runs wellbge-m3-unsupervised BAAI0.6BFP32 (native)maker’s file3.4 / 72.0 GB133–200 tok/s8KUnified memory
Runs wellbge-large-en-v1.5 BAAI0.3BFP32 (native)maker’s file1.8 / 72.0 GB294–442 tok/s512Unified memory
Runs wellbge-reranker-base BAAI0.3BFP32 (native)maker’s file1.6 / 72.0 GB362–543 tok/s514Unified memory
Runs wellbge-base-en-v1.5 BAAI0.1BFP32 (native)maker’s file0.9 / 72.0 GB896–1345 tok/s512Unified memory
Runs wellbge-small-en-v1.5 BAAI0BFP32 (native)maker’s file0.6 / 72.0 GB2866–4298 tok/s512Unified memory
Runs, slowlyQwen2.5-VL-72B-Instruct Alibaba Qwen73.4BQ6_K61.5 / 72.0 GB7–10 tok/s42KUnified memory
Runs, slowlyQwen2-VL-72B-Instruct Alibaba Qwen73.4BQ6_K61.5 / 72.0 GB7–10 tok/s32KUnified memory
Runs, slowlyQwen2.5-72B-Instruct Alibaba Qwen72.7BQ6_Kmaker’s file61.0 / 72.0 GB7–10 tok/s32KUnified memory
Runs, slowlyQianfan-VL-70B Baidu71.8BQ6_K60.2 / 72.0 GB7–10 tok/s32KUnified memory
Runs, slowlyDeepSeek-R1-Distill-Llama-70B DeepSeek70.6BQ6_K59.2 / 72.0 GB7–10 tok/s49KUnified memory

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