β All modelsBROWSE BY RAM
Best local LLMs for 16 GB RAM
With 16 GB of total memory, 38 of the 73 models in our catalog run comfortably at the recommended 4-bit quantization. Top pick: Phi-4 Reasoning Vision 15B (9.1 GB download).
Reading your hardware signalsβ¦
Top picks
All models that run on 16 GB
38 / 73 models| Model | Params | Download (Q4) | Min RAM | Best for |
|---|---|---|---|---|
| Phi-4 Reasoning Vision 15B | 15B | 9.1 GB | 16 GB | Vision, Reasoning |
| Qwen 3 14B | 14.8B | 9.0 GB | 16 GB | Chat, Reasoning |
| DeepSeek R1 14B | 14.8B | 9.0 GB | 16 GB | Reasoning |
| Phi-4 14B | 14.7B | 8.9 GB | 16 GB | Chat, Reasoning |
| Ministral 3 14B | 14B | 8.5 GB | 16 GB | Chat, Vision |
| OLMo 2 13B | 13.7B | 8.3 GB | 12 GB | Chat |
| Gemma 3 12B | 12.2B | 7.4 GB | 12 GB | Chat, Vision |
| Mistral Nemo 12B | 12.2B | 7.4 GB | 12 GB | Chat |
| Gemma 4 12B | 12B | 7.3 GB | 12 GB | Chat, Coding, Reasoning, Vision |
| Mellum 2 12B-A2.5B | 12B (A2.5B) | 7.3 GB | 12 GB | Coding |
| Qwen 3.5 9B | 9B | 5.5 GB | 12 GB | Chat, Reasoning, Vision |
| GLM-4.6V-Flash | 9B | 5.5 GB | 12 GB | Vision, Chat |
| Qwen 2.5 VL 7B | 8.3B | 5.0 GB | 8 GB | Vision, Chat |
| Qwen 3 8B | 8.2B | 5.0 GB | 8 GB | Chat, Reasoning |
| Granite 3.3 8B | 8.2B | 5.0 GB | 8 GB | Chat |
| Llama 3.1 8B | 8B | 4.9 GB | 8 GB | Chat |
| DeepSeek R1 8B | 8B | 4.9 GB | 8 GB | Reasoning |
| Gemma 4 E4B | 8B (A4.5B) | 4.9 GB | 8 GB | Chat, Vision |
| Qwen3-VL 8B | 8B | 4.9 GB | 8 GB | Vision, Chat |
| Ministral 3 8B | 8B | 4.9 GB | 8 GB | Chat, Vision |
| Gemma 3n E4B | 7.8B (A4B) | 4.7 GB | 8 GB | Chat, Vision |
| Qwen 2.5 Coder 7B | 7.6B | 4.6 GB | 8 GB | Coding |
| DeepSeek R1 7B | 7.6B | 4.6 GB | 8 GB | Reasoning |
| Mistral 7B | 7.2B | 4.4 GB | 8 GB | Chat |
| Gemma 4 E2B | 5.1B (A2.3B) | 3.1 GB | 6 GB | Chat, Vision |
| Gemma 3 4B | 4.3B | 2.6 GB | 6 GB | Chat, Vision |
| Qwen 3 4B | 4B | 2.4 GB | 6 GB | Chat, Reasoning |
| Qwen 3.5 4B | 4B | 2.4 GB | 6 GB | Chat, Vision |
| Phi-4 Mini 3.8B | 3.8B | 2.3 GB | 6 GB | Chat |
| Llama 3.2 3B | 3.2B | 1.9 GB | 4 GB | Chat |
| DeepSeek-OCR | 3B (A0.57B) | 1.8 GB | 4 GB | Vision |
| Ministral 3 3B | 3B | 1.8 GB | 4 GB | Chat, Vision |
| DeepSeek R1 1.5B | 1.8B | 1.1 GB | 3 GB | Reasoning |
| Qwen 3 1.7B | 1.7B | 1.0 GB | 3 GB | Chat |
| SmolLM2 1.7B | 1.7B | 1.0 GB | 3 GB | Chat |
| Llama 3.2 1B | 1.2B | 0.7 GB | 3 GB | Chat |
| Gemma 3 1B | 1B | 0.6 GB | 3 GB | Chat |
| Qwen 3 0.6B | 0.6B | 0.4 GB | 2 GB | Chat |
Sizes are 4-bit (Q4_K_M) GGUF builds β the standard for running models locally. Β· Data updated: 2026-06-11 Β· How we calculate these numbers β
Just out of reach β needs 32 GB
These models don't fit comfortably in 16 GB but run on the next tier up.
GPT-OSS 20B Β· 24 GBCodestral 22B Β· 24 GBMistral Small 3.1 24B Β· 24 GBDevstral 24B Β· 24 GBMagistral Small 1.2 Β· 24 GBDevstral Small 2 24B Β· 24 GBGemma 4 26B A4B Β· 24 GBQwen 3.5 27B Β· 24 GBQwen 3.6 27B Β· 24 GBGemma 3 27B Β· 24 GBQwen3-VL 30B-A3B Β· 32 GBQwen 3 30B-A3B Β· 32 GBGemma 4 31B Β· 32 GBNemotron 3 Nano 30B-A3B Β· 32 GBGranite 4.0 H Small Β· 32 GBQwen 3 32B Β· 32 GBQwen 2.5 Coder 32B Β· 32 GBQwQ 32B Β· 32 GBDeepSeek R1 32B Β· 32 GBQwen3-VL 32B Β· 32 GBEXAONE 4.5 33B Β· 32 GBQwen 3.5 35B-A3B Β· 32 GBQwen 3.6 35B-A3B Β· 32 GBCommand R 35B Β· 32 GB
See the 32 GB tier β