Models that run on 256 GB

1161 models with published weights that fit in 256 GB — a Mac Studio with 256. Counted at four-bit quantisation, weights only, leaving the machine about a third of its memory and a gigabyte for context: room for roughly 274 billion parameters. A mixture of experts is counted in full, because it is held in full even though only a few experts compute. 1,161 in all, a hundred to a page; this is page 1 of 12.

A Mac Studio at the top of its configuration, or a serious machine. Almost everything openly published fits, including the large mixtures of experts. At this point the constraint is no longer whether the model loads but whether it generates fast enough to be worth waiting for.

• Inkling SmallThinking Machines Lab266B≈172.9 GB at 4-bit8 also selling it hosted
• Llama-3.1-Nemotron-Ultra-253B-v1NVIDIA253B≈164.5 GB at 4-bit
• Solar-Open2-250BUpstage250B≈162.7 GB at 4-bit
• Intern-S1internlm241B≈156.5 GB at 4-bit
• K-EXAONE-236B-A23BLG AI Research237B≈154.1 GB at 4-bit
• DeepSeek-Coder-V2-InstructDeepSeek236B≈153.2 GB at 4-bit1 also selling it hosted
• DeepSeek-V2DeepSeek236B≈153.2 GB at 4-bit
• DeepSeek-V2-ChatDeepSeek236B≈153.2 GB at 4-bit
• DeepSeek-V2.5DeepSeek236B≈153.2 GB at 4-bit1 also selling it hosted
• DeepSeek-V2.5-1210deepseek-ai236B≈153.2 GB at 4-bit
• Qwen3 VL 235B A22B InstructAlibaba236B≈153.2 GB at 4-bit10 also selling it hosted
• Qwen3 VL 235B A22B ThinkingAlibaba236B≈153.2 GB at 4-bit9 also selling it hosted
• Qwen3 235B A22BAlibaba235B≈152.8 GB at 4-bit10 also selling it hosted
• MiroThinker-v1.5-235BMiroMind AI235B≈152.8 GB at 4-bit
• Qwen3 235B A22B Thinking 2507Alibaba235B≈152.8 GB at 4-bit12 also selling it hosted
• Qwen3-235B-A22B-Instruct-2507Alibaba235B≈152.8 GB at 4-bit13 also selling it hosted
• MiniMax M2.5MiniMax229B≈148.7 GB at 4-bit19 also selling it hosted
• MiniMax M2.1MiniMax229B≈148.6 GB at 4-bit11 also selling it hosted
• MiniMax M2.7MiniMax229B≈148.6 GB at 4-bit14 also selling it hosted
• MiniMax-M2MiniMax229B≈148.6 GB at 4-bit12 also selling it hosted
• Step 3.7 FlashStepFun201B≈130.9 GB at 4-bit9 also selling it hosted
• Step 3.5 FlashStepFun199B≈129.6 GB at 4-bit8 also selling it hosted
• falcon-180BTechnology Innovation Institute180B≈116.7 GB at 4-bit
• falcon-180B-chatTechnology Innovation Institute180B≈116.7 GB at 4-bit
• bloomzBigScience Workshop176B≈114.6 GB at 4-bit
• zephyr-orpo-141b-A35b-v0.1Hugging Face H4141B≈91.7 GB at 4-bit
• Mixtral-8x22B-v0.1Unofficial Mistral Community [deprecated]141B≈91.4 GB at 4-bit
• WizardLM-2 8x22BMicrosoft141B≈91.4 GB at 4-bit6 also selling it hosted
• Ling-3.0-flashInclusionAI127B≈82.9 GB at 4-bit7 also selling it hosted
• Qwen3.5-122B-A10BAlibaba125B≈81.3 GB at 4-bit8 also selling it hosted
• Nemotron 3 SuperNVIDIA124B≈80.3 GB at 4-bit7 also selling it hosted
• Meta-Llama-3-120B-Instructmlabonne122B≈79.2 GB at 4-bit
• GPT OSS Safeguard 120BOpenAI120B≈78.0 GB at 4-bit3 also selling it hosted
• NVIDIA-Nemotron-3-Super-120B-A12B-FP8NVIDIA120B≈78.0 GB at 4-bit
• galactica-120bfacebook120B≈78.0 GB at 4-bit
• goliath-120balpindale118B≈76.5 GB at 4-bit
• Laguna S 2.1Poolside118B≈76.4 GB at 4-bit4 also selling it hosted
• GPT OSS 120BOpenAI117B≈75.9 GB at 4-bit29 also selling it hosted
• c4ai-command-a-03-2025Cohere Labs111B≈72.2 GB at 4-bit
• GLM-4.5-AirZ.ai110B≈71.8 GB at 4-bit12 also selling it hosted
• Llama 4 Scout 17BMeta109B≈70.6 GB at 4-bit7 also selling it hosted
• GLM-4.5VZ.ai108B≈70.0 GB at 4-bit9 also selling it hosted
• GLM-4.6VZ.ai108B≈70.0 GB at 4-bit7 also selling it hosted
• sarvam-105bSarvam AI105B≈68.2 GB at 4-bit
• c4ai-command-r-plusCohere Labs104B≈67.5 GB at 4-bit
• Solar-Open-100BUpstage103B≈66.7 GB at 4-bit
• Llama-3.2-90B-Vision-Instruct90.0B≈58.5 GB at 4-bit7 also selling it hosted
• HunyuanImage-3.0Tencent Hunyuan83.0B≈54.0 GB at 4-bit
• HunyuanImage-3.0-InstructTencent Hunyuan83.0B≈54.0 GB at 4-bit
• Qwen3 Next 80B A3B InstructAlibaba81.3B≈52.9 GB at 4-bit12 also selling it hosted
• Qwen3 Next 80B A3B ThinkingAlibaba81.3B≈52.9 GB at 4-bit9 also selling it hosted
• idefics-80b-instructHuggingFaceM480.0B≈52.0 GB at 4-bit
• Qwen3 Coder NextAlibaba79.7B≈51.8 GB at 4-bit7 also selling it hosted
• NVLM-D-72BNVIDIA79.4B≈51.6 GB at 4-bit
• InternVL3-78BOpenGVLab78.4B≈51.0 GB at 4-bit1 also selling it hosted
• calme-3.2-instruct-78bMaziyarPanahi78.0B≈50.7 GB at 4-bit
• LongCat-NextMeituan LongCat74.3B≈48.3 GB at 4-bit
• Qwen2.5 VL 72B InstructAlibaba73.4B≈47.7 GB at 4-bit7 also selling it hosted
• Kimi-Dev-72BMoonshot AI72.7B≈47.3 GB at 4-bit
• Magnum v4 72BAnthracite72.7B≈47.3 GB at 4-bit1 also selling it hosted
• Qwen2-72BAlibaba72.7B≈47.3 GB at 4-bit
• Qwen2.5 72B InstructAlibaba72.7B≈47.3 GB at 4-bit12 also selling it hosted
• Virtuoso LargeArcee AI72.7B≈47.3 GB at 4-bit
• Smaug-72B-v0.1Abacus.AI, Inc.72.3B≈47.0 GB at 4-bit
• Qwen-72BAlibaba72.3B≈47.0 GB at 4-bit
• Qwen1.5-72B-ChatAlibaba72.3B≈47.0 GB at 4-bit1 also selling it hosted
• KAT-Dev-72B-ExpKwaipilot72.0B≈46.8 GB at 4-bit1 also selling it hosted
• Molmo-72B-0924Allen Institute for AI (Ai2)72.0B≈46.8 GB at 4-bit
• QVQ-72B-PreviewAlibaba72.0B≈46.8 GB at 4-bit1 also selling it hosted
• Qwen 2 VL 72B InstructAlibaba72.0B≈46.8 GB at 4-bit4 also selling it hosted
• Qwen-72B-ChatAlibaba72.0B≈46.8 GB at 4-bit
• Qwen2-72B-InstructAlibaba72.0B≈46.8 GB at 4-bit3 also selling it hosted
• UI-TARS-72B-DPOByteDance Seed72.0B≈46.8 GB at 4-bit
• dolphin-2.9.2-qwen2-72bdphn72.0B≈46.8 GB at 4-bit1 also selling it hosted
• magnum-v1-72banthracite-org72.0B≈46.8 GB at 4-bit
• Llama3-ChatQA-1.5-70BNVIDIA70.6B≈45.9 GB at 4-bit
• Athene-70BNexusflow70.6B≈45.9 GB at 4-bit
• Hermes 3 70B InstructNous Research70.6B≈45.9 GB at 4-bit4 also selling it hosted
• Hermes 4 70BNous Research70.6B≈45.9 GB at 4-bit4 also selling it hosted
• Higgs-Llama-3-70BBoson AI70.6B≈45.9 GB at 4-bit
• Llama 3.1 Instruct (70B)Meta70.6B≈45.9 GB at 4-bit7 also selling it hosted
• Llama 3.3 70B InstructMeta70.6B≈45.9 GB at 4-bit29 also selling it hosted
• Llama-3.1-Nemotron-70B-Instruct-HFNVIDIA70.6B≈45.9 GB at 4-bit1 also selling it hosted
• R1 Distill Llama 70BDeepSeek70.6B≈45.9 GB at 4-bit12 also selling it hosted
• Apertus-70B-2509swiss-ai70.0B≈45.5 GB at 4-bit
• Apertus-70B-Instruct-2509swiss-ai70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Apertus-v1.5-70Bswiss-ai70.0B≈45.5 GB at 4-bit1 also selling it hosted
• CodeLlama-70b-hfCode Llama70.0B≈45.5 GB at 4-bit
• L3 70B Euryale V2.1Sao10K70.0B≈45.5 GB at 4-bit3 also selling it hosted
• L3.3-70B-Euryale-v2.3Sao10K70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Llama-2-70bMeta Llama70.0B≈45.5 GB at 4-bit
• Llama-2-70b-chatMeta Llama70.0B≈45.5 GB at 4-bit2 also selling it hosted
• Llama-2-70b-chat-hfMeta70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Llama-3-Groq-70B-Tool-UseGroq70.0B≈45.5 GB at 4-bit
• Llama-3.1-Nemotron-70B-InstructNVIDIA70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Llama-xLAM-2-70b-fc-rSalesforce70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Llama3-OpenBioLLM-70Baaditya70.0B≈45.5 GB at 4-bit
• Meta-Llama-3-70B-InstructMeta70.0B≈45.5 GB at 4-bit6 also selling it hosted
• Meta-Llama-3.1-70B-InstructMeta70.0B≈45.5 GB at 4-bit7 also selling it hosted
• Platypus2-70B-instructgarage-bAInd70.0B≈45.5 GB at 4-bit