Conversation models that run on 256 GB

628 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. 628 in all, a hundred to a page; this is page 6 of 7.

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.

Models that answer in prose across whatever subject you put to them. This is the largest category and the least differentiated: most of them are competent at most things, so the choice usually comes down to price, context length and how the maker behaves about availability. Look at what a model costs per million tokens in and out — the output rate is often four or five times the input rate, and it is the one that decides your bill.

• Ovis2.5-2BATH-MaaS2.6B≈1.7 GB at 4-bit
• LFM2-2.6BLiquid AI2.6B≈1.7 GB at 4-bit
• LFM2-2.6B-ExpLiquid AI2.6B≈1.7 GB at 4-bit
• MiniCPM5-2BOpenBMB2.5B≈1.6 GB at 4-bit
• Octopus-v2Nexa AI2.5B≈1.6 GB at 4-bit
• gemma-2bGoogle2.5B≈1.6 GB at 4-bit
• gemma-2b-itGoogle2.5B≈1.6 GB at 4-bit1 also selling it hosted
• Cosmos-Reason2-2BNVIDIA2.4B≈1.6 GB at 4-bit
• EXAONE-3.5-2.4B-InstructLGAI-EXAONE2.4B≈1.6 GB at 4-bit
• SmolVLM2-2.2B-InstructHugging Face Smol Models Research2.2B≈1.5 GB at 4-bit
• Marlin-2BNemo Station2.2B≈1.4 GB at 4-bit
• Qwen2-VL-2B-InstructAlibaba2.2B≈1.4 GB at 4-bit1 also selling it hosted
• Qwen3-VL-2B-InstructAlibaba2.1B≈1.4 GB at 4-bit
• MiniCPM-2B-sft-fp32OpenBMB2.0B≈1.3 GB at 4-bit
• Qwen3.5-2BAlibaba2.0B≈1.3 GB at 4-bit2 also selling it hosted
• gemma-1.1-2b-itGoogle2.0B≈1.3 GB at 4-bit
• helium-1-preview-2bKyutai2.0B≈1.3 GB at 4-bit
• Youtu-LLM-2BTencent Hunyuan2.0B≈1.3 GB at 4-bit
• moondream2vikhyatk1.9B≈1.3 GB at 4-bit
• VibeThinker-1.5BWeiboAI1.8B≈1.2 GB at 4-bit
• Qwen3.6-27B-DFlashZ Lab1.7B≈1.1 GB at 4-bit
• SmolLM-1.7BHuggingFaceTB1.7B≈1.1 GB at 4-bit
• SmolLM2-1.7BHuggingFaceTB1.7B≈1.1 GB at 4-bit
• stablelm-2-1_6bStability AI1.6B≈1.1 GB at 4-bit
• stablelm-2-zephyr-1_6bStability AI1.6B≈1.1 GB at 4-bit
• LFM2.5-VL-1.6BLiquid AI1.6B≈1.0 GB at 4-bit
• LFM2-VL-1.6BLiquid AI1.6B≈1.0 GB at 4-bit
• Qwen2.5-1.5BAlibaba1.5B≈1.0 GB at 4-bit
• Qwen2.5-1.5B-InstructAlibaba1.5B≈1.0 GB at 4-bit1 also selling it hosted
• ReaderLM-v2Jina AI1.5B≈1.0 GB at 4-bit
• reader-lm-1.5bJina AI1.5B≈1.0 GB at 4-bit
• Hymba-1.5B-InstructNVIDIA1.5B≈1.0 GB at 4-bit
• Arch-Router-1.5Bkatanemo1.5B≈1.0 GB at 4-bit
• Hymba-1.5B-BaseNVIDIA1.5B≈1.0 GB at 4-bit
• HyperCLOVAX-SEED-Text-Instruct-1.5BHyperCLOVA X1.5B≈1.0 GB at 4-bit
• Qwen2-1.5B-InstructAlibaba1.5B≈1.0 GB at 4-bit1 also selling it hosted
• starvector-1b-im2svgstarvector1.4B≈0.9 GB at 4-bit
• phi-1Microsoft1.4B≈0.9 GB at 4-bit
• phi-1_5Microsoft1.4B≈0.9 GB at 4-bit
• Ouro-1.4BByteDance1.4B≈0.9 GB at 4-bit
• MiniCPM-V-4.6OpenBMB1.3B≈0.8 GB at 4-bit
• gpt-neo-1.3BEleutherAI1.3B≈0.8 GB at 4-bit
• nllb-200-distilled-1.3Bfacebook1.3B≈0.8 GB at 4-bit
• opt-1.3bfacebook1.3B≈0.8 GB at 4-bit
• EXAONE-4.0-1.2BLGAI-EXAONE1.3B≈0.8 GB at 4-bit
• SpatialLM-Llama-1BManycore Research1.2B≈0.8 GB at 4-bit
• LFM2.5-1.2B-BaseLiquid AI1.2B≈0.8 GB at 4-bit
• LFM2.5-1.2B-JPLiquid AI1.2B≈0.8 GB at 4-bit
• HRM-Text-1BSapient AI1.2B≈0.8 GB at 4-bit
• LFM2-1.2BLiquid AI1.2B≈0.8 GB at 4-bit
• LFM2.5-1.2B-InstructLiquid AI1.2B≈0.8 GB at 4-bit
• MinerU2.5-2509-1.2BOpenDataLab1.2B≈0.8 GB at 4-bit
• TinyLlama-1.1B-Chat-v1.0TinyLlama1.1B≈0.7 GB at 4-bit
• TinyLlama-1.1B-intermediate-step-1431k-3TTinyLlama1.1B≈0.7 GB at 4-bit
• MiniCPM5-1BOpenBMB1.1B≈0.7 GB at 4-bit
• vaultgemma-1bGoogle1.0B≈0.7 GB at 4-bit
• Gemma3-1B-ITLiteRT Community (FKA TFLite)1.0B≈0.7 GB at 4-bit
• MolmoE-1B-0924Allen Institute for AI (Ai2)1.0B≈0.7 GB at 4-bit
• antares-1bfdtn-ai1.0B≈0.7 GB at 4-bit
• granite-4.0-h-1bIBM Granite1.0B≈0.7 GB at 4-bit
• gemma-3-1b-itGoogle1.0B≈0.6 GB at 4-bit
• gemma-3-1b-ptGoogle1.0B≈0.6 GB at 4-bit
• CLIP-ViT-H-14-laion2B-s32B-b79KLAION eV1.0B≈0.6 GB at 4-bit
• MobileLLM-R1-950MAI at Meta0.9B≈0.6 GB at 4-bit
• siglip-so400m-patch14-384Google0.9B≈0.6 GB at 4-bit
• bitnet-b1.58-2B-4TMicrosoft0.8B≈0.6 GB at 4-bit
• gpt2-largeOpenAI community0.8B≈0.5 GB at 4-bit
• t5gemma-2-270m-270mGoogle0.8B≈0.5 GB at 4-bit
• Florence-2-largeMicrosoft0.8B≈0.5 GB at 4-bit
• Florence-2-large-ftMicrosoft0.8B≈0.5 GB at 4-bit
• FastVLM-0.5BApple0.8B≈0.5 GB at 4-bit
• Qwen3-0.6BAlibaba0.8B≈0.5 GB at 4-bit1 also selling it hosted
• Qwen3-0.6B-BaseAlibaba0.6B≈0.4 GB at 4-bit
• bloom-560mBigScience Workshop0.6B≈0.4 GB at 4-bit
• Qwen1.5-0.5BAlibaba0.5B≈0.3 GB at 4-bit
• Qwen2-0.5BAlibaba0.5B≈0.3 GB at 4-bit
• reader-lm-0.5bJina AI0.5B≈0.3 GB at 4-bit
• Qwen2-0.5B-InstructAlibaba0.5B≈0.3 GB at 4-bit
• Qwen2.5-0.5BAlibaba0.5B≈0.3 GB at 4-bit
• Qwen2.5-0.5B-InstructAlibaba0.5B≈0.3 GB at 4-bit1 also selling it hosted
• blip-image-captioning-largeSalesforce0.5B≈0.3 GB at 4-bit
• LFM2.5-VL-450MLiquid AI0.4B≈0.3 GB at 4-bit
• PairRMLLM Blender0.4B≈0.3 GB at 4-bit
• CLIP-GmP-ViT-L-14zer0int0.4B≈0.3 GB at 4-bit
• circuit-sparsityOpenAI0.4B≈0.3 GB at 4-bit
• bart-large-cnnAI at Meta0.4B≈0.3 GB at 4-bit
• ModernBERT-largeAnswer.AI0.4B≈0.3 GB at 4-bit
• blip-vqa-baseSalesforce0.4B≈0.3 GB at 4-bit
• LFM2-350MLiquid AI0.4B≈0.2 GB at 4-bit
• LFM2.5-350MLiquid AI0.4B≈0.2 GB at 4-bit
• nougat-basefacebook0.3B≈0.2 GB at 4-bit
• bert-large-uncased-whole-word-masking-finetuned-squadBERT community0.3B≈0.2 GB at 4-bit
• mdeberta-v3-base-squad2timpal0l0.3B≈0.2 GB at 4-bit
• functiongemma-270m-itGoogle0.3B≈0.2 GB at 4-bit
• gemma-3-270mGoogle0.3B≈0.2 GB at 4-bit
• gemma-3-270m-itGoogle0.3B≈0.2 GB at 4-bit
• granite-docling-258MIBM Granite0.3B≈0.2 GB at 4-bit
• Florence-2-baseMicrosoft0.2B≈0.2 GB at 4-bit
• t5-baseT5 community0.2B≈0.1 GB at 4-bit
• chatgpt_paraphraser_on_T5_baseHumarin0.2B≈0.1 GB at 4-bit

Wider