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 2 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.

• Reflection-Llama-3.1-70Bmattshumer70.0B≈45.5 GB at 4-bit1 also selling it hosted
• Smaug-Llama-3-70B-InstructAbacus.AI, Inc.70.0B≈45.5 GB at 4-bit
• WizardLM-70B-V1.0WizardLM Team70.0B≈45.5 GB at 4-bit
• Xwin-LM-70B-V0.1Xwin-LM70.0B≈45.5 GB at 4-bit
• dolphin-2.9.1-llama-3-70bcognitivecomputations70.0B≈45.5 GB at 4-bit1 also selling it hosted
• llama-3-70B-Instruct-abliteratedfailspy70.0B≈45.5 GB at 4-bit
• lzlv_70b_fp16_hflizpreciatior70.0B≈45.5 GB at 4-bit1 also selling it hosted
• med42-70bM42 Health70.0B≈45.5 GB at 4-bit
• meditron-70bEPFL LLM Team70.0B≈45.5 GB at 4-bit
• tulu-2-dpo-70bAllen Institute for AI (Ai2)70.0B≈45.5 GB at 4-bit
• LongCat-Flash-LiteMeituan LongCat69.1B≈44.9 GB at 4-bit
• CodeLlama-70b-Instruct-hfCode Llama69.0B≈44.8 GB at 4-bit
• sqlcoder-70b-alphadefog69.0B≈44.8 GB at 4-bit
• Llama-2-70b-hfMeta Llama69.0B≈44.8 GB at 4-bit
• NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4NVIDIA67.2B≈43.7 GB at 4-bit
• deepseek-llm-67b-chatdeepseek-ai67.0B≈43.6 GB at 4-bit
• opt-66bfacebook66.0B≈42.9 GB at 4-bit
• Jamba-v0.1AI21 Labs51.6B≈33.5 GB at 4-bit
• Llama-3_1-Nemotron-51B-InstructNVIDIA51.5B≈33.5 GB at 4-bit
• Kimi-Linear-48B-A3B-InstructMoonshot AI49.1B≈31.9 GB at 4-bit
• Llama-3_3-Nemotron-Super-49B-v1NVIDIA49.0B≈31.9 GB at 4-bit
• dolphin-2.5-mixtral-8x7bDolphin46.7B≈30.4 GB at 4-bit
• GRIN-MoEMicrosoft41.9B≈27.2 GB at 4-bit
• Phi-3.5-MoE-instructMicrosoft41.9B≈27.2 GB at 4-bit1 also selling it hosted
• falcon-40bTechnology Innovation Institute41.8B≈27.2 GB at 4-bit
• IQuest-Coder-V1-40B-InstructIQuest40.0B≈26.0 GB at 4-bit
• falcon-40b-instructTechnology Innovation Institute40.0B≈26.0 GB at 4-bit
• IQuest-Coder-V1-40B-Loop-InstructIQuest39.8B≈25.9 GB at 4-bit
• Qwen3.5-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-ThinkingDavidAU39.5B≈25.7 GB at 4-bit
• CLIP-ViT-bigG-14-laion2B-39B-b160kLAION eV39.0B≈25.4 GB at 4-bit
• Seed-OSS-36B-InstructByteDance Seed36.2B≈23.5 GB at 4-bit1 also selling it hosted
• K2-Horizon-MoVA-36B-A4BInstitute of Foundation Models36.0B≈23.4 GB at 4-bit
• Skyfall 36B V2TheDrummer36.0B≈23.4 GB at 4-bit1 also selling it hosted
• BigBang-v1The Endless Frontier36.0B≈23.4 GB at 4-bit
• Huihui-Qwen3.5-35B-A3B-abliteratedhuihui-ai36.0B≈23.4 GB at 4-bit
• Ornith-1.5-35B-A3BOrnith36.0B≈23.4 GB at 4-bit
• Qwable-v1lordx6436.0B≈23.4 GB at 4-bit
• Qwen3.5-35B-A3BAlibaba36.0B≈23.4 GB at 4-bit9 also selling it hosted
• Qwen3.6 35B A3BAlibaba36.0B≈23.4 GB at 4-bit10 also selling it hosted
• Agents-A1Intern Science35.1B≈22.8 GB at 4-bit
• Nex-N2-MiniNEX AGI35.1B≈22.8 GB at 4-bit
• Nex-N2.5-miniNEX AGI35.1B≈22.8 GB at 4-bit1 also selling it hosted
• Thomson-1.0-SmallThomson Reuters35.1B≈22.8 GB at 4-bit
• XYZ-Aquila-miniXYZAILab35.1B≈22.8 GB at 4-bit
• Holo3-35B-A3BH Company35.0B≈22.8 GB at 4-bit1 also selling it hosted
• Ornith-1.0-35Bdeepreinforce-ai35.0B≈22.8 GB at 4-bit1 also selling it hosted
• Qwen3.5-35B-A3B-BaseAlibaba35.0B≈22.8 GB at 4-bit1 also selling it hosted
• Qwen3.6-35B-A3B-DFlashZ Lab35.0B≈22.8 GB at 4-bit
• aya-23-35BCohere Labs35.0B≈22.8 GB at 4-bit
• c4ai-command-r-v01Cohere Labs35.0B≈22.7 GB at 4-bit
• llava-v1.6-34bliuhaotian34.8B≈22.6 GB at 4-bit
• KAT-Coder-V2.5-DevKwaipilot34.7B≈22.5 GB at 4-bit
• Qwen-AgentWorld-35B-A3BAlibaba34.7B≈22.5 GB at 4-bit
• Yi-34B01-ai34.4B≈22.4 GB at 4-bit1 also selling it hosted
• Yi-34B-Chat01-ai34.4B≈22.4 GB at 4-bit1 also selling it hosted
• CodeBooga-34B-v0.1oobabooga34.0B≈22.1 GB at 4-bit
• CodeLlama-34b-Instruct-hfCode Llama34.0B≈22.1 GB at 4-bit
• CodeLlama-34b-hfCode Llama34.0B≈22.1 GB at 4-bit
• Ovis2-34BATH-MaaS34.0B≈22.1 GB at 4-bit
• Phind-CodeLlama-34B-Python-v1Phind34.0B≈22.1 GB at 4-bit1 also selling it hosted
• Phind-CodeLlama-34B-v1Phind34.0B≈22.1 GB at 4-bit1 also selling it hosted
• Phind-CodeLlama-34B-v2Phind34.0B≈22.1 GB at 4-bit2 also selling it hosted
• WizardCoder-Python-34B-V1.0WizardLM Team34.0B≈22.1 GB at 4-bit
• Yi-1.5-34B-Chat01-ai34.0B≈22.1 GB at 4-bit
• Yi-VL-34B01-ai34.0B≈22.1 GB at 4-bit
• deepsex-34bTriadParty34.0B≈22.1 GB at 4-bit
• sqlcoder-34b-alphadefog34.0B≈22.1 GB at 4-bit
• Laguna XS 2.1Poolside33.4B≈21.7 GB at 4-bit3 also selling it hosted
• Qwen3 VL 32B InstructAlibaba33.4B≈21.7 GB at 4-bit5 also selling it hosted
• deepseek-coder-33b-instructDeepSeek33.3B≈21.7 GB at 4-bit2 also selling it hosted
• HyperCLOVAX-SEED-Think-32BHyperCLOVA X33.3B≈21.7 GB at 4-bit
• aya-vision-32bCohere Labs33.1B≈21.5 GB at 4-bit
• Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16NVIDIA33.0B≈21.5 GB at 4-bit
• EXAONE-4.5-33BLGAI-EXAONE33.0B≈21.4 GB at 4-bit
• OpenCodeInterpreter-DS-33BMultimodal Art Projection33.0B≈21.4 GB at 4-bit
• Qwen3 32BAlibaba32.8B≈21.3 GB at 4-bit17 also selling it hosted
• AM-Thinking-v1am team32.8B≈21.3 GB at 4-bit
• DeepSeek-R1-Distill-Qwen-32B-JapaneseCyberAgent32.8B≈21.3 GB at 4-bit
• K2-ThinkInstitute of Foundation Models32.8B≈21.3 GB at 4-bit
• Qwen2.5 Coder 32B InstructAlibaba32.8B≈21.3 GB at 4-bit11 also selling it hosted
• TinyR1-32B-Previewqihoo36032.8B≈21.3 GB at 4-bit
• openhands-lm-32b-v0.1OpenHands32.8B≈21.3 GB at 4-bit
• DeepSWE-PreviewAgentica32.8B≈21.3 GB at 4-bit
• KAT-DevKwaipilot32.8B≈21.3 GB at 4-bit
• GLM-Z1-32B-0414zai-org32.6B≈21.2 GB at 4-bit
• Olmo 3 32B ThinkAllen Institute for AI (Ai2)32.2B≈21.0 GB at 4-bit
• FLUX.2 [dev]Black Forest Labs32.2B≈20.9 GB at 4-bit1 also selling it hosted
• sarvam-30bSarvam AI32.2B≈20.9 GB at 4-bit
• Baichuan M2 32Bbaichuan32.0B≈20.8 GB at 4-bit2 also selling it hosted
• DeepSeek R1 Distill QWEN 32BDeepSeek32.0B≈20.8 GB at 4-bit6 also selling it hosted
• EXAONE-Deep-32BLGAI-EXAONE32.0B≈20.8 GB at 4-bit
• GLM-4-32B-0414Z.ai32.0B≈20.8 GB at 4-bit3 also selling it hosted
• Olmo-2-0325-32B-InstructAllen Institute for AI (Ai2)32.0B≈20.8 GB at 4-bit
• Olmo-3.1-32B-InstructAllen Institute for AI (Ai2)32.0B≈20.8 GB at 4-bit1 also selling it hosted
• OlympicCoder-32Bopen-r132.0B≈20.8 GB at 4-bit
• OpenReasoning-Nemotron-32BNVIDIA32.0B≈20.8 GB at 4-bit1 also selling it hosted
• OpenThinker-32Bopen-thoughts32.0B≈20.8 GB at 4-bit
• QwQ-32BAlibaba32.0B≈20.8 GB at 4-bit7 also selling it hosted
• QwQ-32B-PreviewAlibaba32.0B≈20.8 GB at 4-bit2 also selling it hosted
• QwQ-R1984-32BVIDraft32.0B≈20.8 GB at 4-bit