// Open-model VRAM table · family
Ai2 OLMo / Molmo: local deployment & VRAM
31 open-weight models from Allen Institute for AI (52 official repos incl. quantised builds). At 4-bit they range from 2.4 GB (OLMo-1B-0724-hf) to 672 GB (FlexOlmo-7x7B-1T).
official orgs: allenailatest 2026-05-04memory = estimate
01Text LLM (19)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| Olmo-Hybrid-7B | 7.43B | 64K | 17.9 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-01-28 |
| Olmo-3.1-32B-Instruct | 32.2B | 64K | 72.8 GB | 23.3 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2025-12-10 |
| Olmo-3.1-32B-Think | 32.2B | 64K | 72.8 GB | 23.3 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2025-12-10 |
| Olmo-3-7B-Instruct | 7.3B | 64K | 19.2 GB | 8.0 GB | 12GB GPU | Mac 16GB | Commercial OK | 2025-11-19 |
| Olmo-3-32B-Think | 32.2B | 64K | 72.8 GB | 23.3 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2025-11-19 |
| Olmo-3-7B-Think | 7.3B | 64K | 19.2 GB | 8.0 GB | 12GB GPU | Mac 16GB | Commercial OK | 2025-11-18 |
| Olmo-3-1125-32B | 32.2B | 64K | 72.8 GB | 23.3 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2025-11-04 |
| Olmo-3-1025-7B | 7.3B | 64K | 19.2 GB | 8.0 GB | 12GB GPU | Mac 16GB | Commercial OK | 2025-09-12 |
| FlexOlmo-7x7B-1T | 1T | 4K | 2.20 TB | 672 GB | 8 × H200 (1,128GB) | — | Commercial OK | 2025-06-11 |
| OLMo-2-0425-1B | 1.49B | 4K | 4.8 GB | 2.6 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-04-17 |
| OLMo-2-0325-32B-Instruct | 32.2B | 4K | 73.6 GB | 24.1 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2025-03-12 |
| OLMoE-1B-7B-0125-Instruct | 6.92B | 4K | 16.8 GB | 6.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-01-27 |
| Llama-3.1-Tulu-3-405B | 405.9B | 128K | 898 GB | 275 GB | 4 × RTX PRO 6000 (384GB) | Mac 512GB | Commercial with conditions | 2025-01-09 |
| Llama-3.1-Tulu-3-8B | 8.03B | 128K | 19.2 GB | 6.9 GB | 8GB GPU | Mac 16GB | Commercial with conditions | 2024-11-20 |
| OLMo-2-1124-13B | 13.7B | 4K | 37.4 GB | 16.4 GB | RTX 4090 24GB | Mac 36GB | Commercial OK | 2024-11-19 |
| OLMo-2-1124-7B | 7.3B | 4K | 20.8 GB | 9.7 GB | 12GB GPU | Mac 24GB | Commercial OK | 2024-10-29 |
| OLMoE-1B-7B-0924 | 6.92B | 4K | 16.8 GB | 6.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-07-20 |
| OLMo-7B-0724-hf | 6.89B | 4K | 19.9 GB | 9.4 GB | 12GB GPU | Mac 16GB | Commercial OK | 2024-07-12 |
| OLMo-1B-0724-hf | 1.28B | 4K | 4.4 GB | 2.4 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-06-15 |
02Vision-language (12)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| Molmo2-ER | 4.85B | 16K | 12.4 GB | 4.9 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-05-04 |
| MolmoWeb-4B ↗ | 4.85B | 10K | 12.4 GB | 4.9 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-03-20 |
| MolmoPoint-8B | 8.68B | 37K | 20.8 GB | 7.5 GB | 12GB GPU | Mac 16GB | Commercial OK | 2026-03-16 |
| Molmo2-4B | 4.85B | 36K | 12.4 GB | 4.9 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-12-14 |
| Molmo2-8B | 8.66B | 36K | 20.8 GB | 7.5 GB | 12GB GPU | Mac 16GB | Commercial OK | 2025-12-14 |
| Molmo2-O-7B | 7.76B | 64K | 21.9 GB | 10.0 GB | 12GB GPU | Mac 24GB | Commercial OK | 2025-12-14 |
| olmOCR-2-7B-1025 | 8.29B | 125K | 19.2 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-10-06 |
| olmOCR-7B-0225-preview | 8.29B | 32K | 19.2 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-01-15 |
| Molmo-72B-0924 | 73.3B | 4K | 165 GB | 52.1 GB | 2 × RTX 5090 (64GB) | Mac 128GB | Commercial OK | 2024-09-25 |
| Molmo-7B-D-0924 | 8.02B | 4K | 18.6 GB | 6.3 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-09-25 |
| Molmo-7B-O-0924 ↗ | 7.67B | 4K | 21.7 GB | 9.9 GB | 12GB GPU | Mac 24GB | Commercial OK | 2024-09-25 |
| MolmoE-1B-0924 ↗ | 7.2B / 1.5B | 32K | 16.3 GB | 5.3 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-09-24 |
03Other families
Qwen · IBM Granite · NVIDIA Nemotron · Gemma · Mistral / Magistral / Devstral · GLM / CogVideoX (Z.ai) · DeepSeek · Ant Ling / Ring / Ming · Liquid LFM · MiniCPM / VoxCPM · Tencent Hunyuan / Hy · Phi · Llama · Meituan LongCat · Jina embeddings & rerankers · Wan · Kimi · Cohere Command / Aya · Xiaomi MiMo · FLUX · Microsoft VibeVoice / Fara / Harrier · InternLM / Intern-S · StepFun Step · MiniMax · Baidu ERNIE · NVIDIA Parakeet / Canary speech · Stable Diffusion / Stable Audio · Yi (01.AI) · ByteDance Seed · SmolLM / SmolVLM · BGE embeddings & rerankers · LTX Video · FunAudioLLM speech · gpt-oss · poolside Laguna · Whisper · Baichuan · Thinking Machines Inkling · xAI Grok · rednote dots
Totals include 10% + 0.5GB overhead and one 8K-token KV cache for text models. Estimates only. Method →