Skip to content
localmodel.run

Memory budget · 16 GB

Best local LLMs for 16GB

16GB is not a single ceiling. A 16GB Mac, a 16GB GPU and a 16GB Laptop each leave a different amount free for model weights, so the largest model you can run changes with the memory type, not just the number.

Usable range
10.5–15 GB
Models that fit
69
Memory types
3
Top pick
21B

What 16GB actually gives you

Apple unified memory ~10.5 GB usable on Apple M5 (16GB) Comfortable ceiling: 12.2B GPU VRAM ~15 GB usable on Nvidia GeForce RTX 4080 (16GB) Comfortable ceiling: 21B System RAM (CPU only) ~12 GB usable on 16GB RAM Laptop (CPU/iGPU only) Comfortable ceiling: 14B

Usable figures are sourced per device (tap a card for the full profile). Verdicts below use Q4_K_M, the community-default quant.

Top pick for 16GB Q4_K_M
gpt-oss 20B 21B MoE

Runs comfortably on the most capable 16GB setup (Nvidia GeForce RTX 4080 (16GB), ~15 GB usable) at ~13.2 GB. Check it against your exact device on its model page.

Models ranked for 16GB

Biggest that fits first Mac · GPU · Laptop

Each chip links to the full breakdown for that model on a real 16GB device. "Tight" means it fits but with little headroom, close other apps.

The ceiling, per memory type

Apple M5 (16GB) (~10.5 GB usable)

Runs up to Mistral Nemo 12B (12.2B) comfortably at Q4_K_M. Larger models either sit tight or spill past the ~10.5 GB it can give a model.

Nvidia GeForce RTX 4080 (16GB) (~15 GB usable)

Runs up to gpt-oss 20B (21B) comfortably at Q4_K_M. Larger models either sit tight or spill past the ~15 GB it can give a model.

16GB RAM Laptop (CPU/iGPU only) (~12 GB usable)

Runs up to Phi-4-reasoning (14B) comfortably at Q4_K_M. Larger models either sit tight or spill past the ~12 GB it can give a model.

16GB phones & tablets

Phones report 16GB too, but iOS/Android reserve more and the runtimes differ. Their usable pool is smaller:

iPad Pro M4 (16GB, 1TB/2TB config) ~12 GB Google Pixel 9 Pro ~10.5 GB Samsung Galaxy S25 Ultra (16GB, 1TB config only) ~12 GB Google Pixel 10 Pro ~10.5 GB Samsung Galaxy S26 Ultra (16GB, 1TB config) ~12 GB

Too large for any 16GB device

Mistral Small 3 24B 24B Sarvam-M 24B 24B Mistral Small 3.1 24B 24B Magistral Small 24B Devstral Small 24B LFM2 24B-A2B 24B Gemma 4 26B-A4B 26.5B Gemma 2 27B 27B Gemma 3 27B 27B Qwen3.6 27B 27.8B Granite 4.1 30B 28.9B Sarvam-30B 30B Nemotron 3 Nano 30B-A3B 30B Nemotron Cascade 2 30B-A3B 30B GLM-4.7-Flash 30B Qwen3 30B-A3B 30.5B Qwen3-Coder 30B-A3B 30.5B North Mini Code 1.0 30.5B Qwen2.5 32B 32B Qwen3 32B 32B DeepSeek-R1-Distill-Qwen 32B 32B Qwen2.5 Coder 32B 32B Granite 4.0 H Small 32B GLM-4-32B-0414 32B EXAONE 4.0 32B 32B OLMo 2 32B Instruct 32B Granite 4.0 H Small 32B Olmo 3.1 32B Instruct 32B Gemma 4 31B 32.7B Laguna XS 2.1 33.4B Yi 1.5 34B 34B Falcon-H1-34B-Instruct 34B Qwen-AgentWorld 35B-A3B 34.7B Command R 35B 35B Ornith 1.0 35B 35B Seed-OSS 36B Instruct 36B Qwen3.6 35B-A3B 36B Mixtral 8x7B 46.7B Llama-3.3-Nemotron-Super-49B-v1 49B Llama 3.3 70B 70B Qwen2.5 72B 72B Hunyuan-A13B-Instruct 80B Sarvam-105B 105B GLM-4.5-Air 106B Llama 4 Scout 109B Command A 111B gpt-oss 120B 117B Laguna S 2.1 118B Nemotron 3 Super 120B-A12B 120B dots.llm1 142B Qwen3 235B A22B 235B DeepSeek-V4-Flash 284B GLM-4.6 357B Llama 4 Maverick 400B MiniMax M3 428B MiniMax-M1-80k 456B Qwen3-Coder 480B-A35B Instruct 480B Nemotron 3 Ultra 550B-A55B 550B DeepSeek R1 671B DeepSeek V3 671B DeepSeek-R1-0528 671B GLM-5.2 744B Kimi K2 Instruct 1000B Kimi K2.6 1000B Kimi K2.7 Code 1000B DeepSeek-V4-Pro 1600B

FAQ

How much of 16GB can a model actually use?

It depends on the memory type. Apple unified memory: about 10.5 GB (Apple M5 (16GB)); GPU VRAM: about 15 GB (Nvidia GeForce RTX 4080 (16GB)); System RAM (CPU only): about 12 GB (16GB RAM Laptop (CPU/iGPU only)). The rest is reserved for the OS, display and runtime overhead.

What is the best local LLM for 16GB?

gpt-oss 20B (21B) is the strongest model that runs comfortably at Q4_K_M on the most capable 16GB setup (Nvidia GeForce RTX 4080 (16GB), ~15 GB usable). On a tighter 16GB device the ceiling is lower, shown per row above.

Why does a 16GB GPU fit a bigger model than a 16GB Mac?

A discrete GPU gives almost all of its VRAM to the model (leave ~1 GB for the driver). Apple Silicon shares one unified pool with macOS, so roughly 66% is available to the GPU for weights. Same 16GB sticker, different usable budget, so the model ceiling differs.

Sources

Memory figures are estimates at Q4_K_M with a small context. See methodology.