Best LLMs for 16GB VRAM
With 16 GB of VRAM you can run models up to Gemma 4 26B-A4B (25.2B) at Q4_K_M with an 8K context. 28 of 58 tracked models fit fully, and 13 more fit at a lower quantization.
Models that fit in 16 GB
| Model | VRAM at 8K | Best quant that fits | Max context at Q4 |
|---|---|---|---|
| Gemma 4 26B-A4B | 15.6 GB | IQ4_XS | – |
| Mistral Small 3.1 24B | 15.6 GB | IQ4_XS | 4K |
| gpt-oss-20b | 13.9 GB | MXFP4 (official) | 55K |
| Qwen 3 14B | 10.5 GB | Q6_K | 37K |
| Qwen 2.5 14B | 10.7 GB | Q6_K | 31K |
| Phi-4 14B | 10.6 GB | Q6_K | 16K |
| DeepSeek R1 Distill Qwen 14B | 10.7 GB | Q6_K | 31K |
| Ministral 3 14B | 9.7 GB | Q6_K | 42K |
| Mistral Nemo 12B | 9.0 GB | Q8_0 | 46K |
| Gemma 3 12B | 8.8 GB | Q8_0 | 108K |
| Qwen 3.5 9B | 6.4 GB | Q8_0 | 256K |
| Ministral 3 8B | 6.7 GB | Q8_0 | 71K |
| Qwen 3 8B | 6.6 GB | Q8_0 | 40K |
| Llama 3.1 8B | 6.3 GB | Q8_0 | 77K |
| DeepSeek R1 Distill Llama 8B | 6.3 GB | Q8_0 | 77K |
| Gemma 4 E4B | 5.7 GB | Q8_0 | 128K |
| Qwen 2.5 7B | 5.6 GB | Q8_0 | 32K |
| DeepSeek R1 Distill Qwen 7B | 5.6 GB | Q8_0 | 128K |
| Mistral 7B v0.3 | 5.7 GB | Q8_0 | 32K |
| Gemma 4 E2B | 4.0 GB | FP16 / BF16 | 128K |
| Gemma 3 4B | 3.7 GB | FP16 / BF16 | 128K |
| Qwen 3.5 4B | 3.7 GB | FP16 / BF16 | 256K |
| Qwen 3 4B | 4.2 GB | FP16 / BF16 | 40K |
| Phi-4 Mini | 4.1 GB | FP16 / BF16 | 95K |
| Ministral 3 3B | 3.5 GB | FP16 / BF16 | 121K |
| Llama 3.2 3B | 3.5 GB | FP16 / BF16 | 113K |
| Llama 3.2 1B | 1.7 GB | FP16 / BF16 | 128K |
| Gemma 3 1B | 1.6 GB | FP16 / BF16 | 32K |
Fits at a lower quantization
These need more than 16 GB at Q4_K_M but fit at a smaller format, with more quality loss.
| Model | VRAM at Q4_K_M | Fits at |
|---|---|---|
| Qwen 3.5 35B-A3B | 21.0 GB | Q2_K |
| Qwen 3.6 35B-A3B | 21.0 GB | Q2_K |
| Qwen 2.5 32B | 21.4 GB | Q2_K |
| Qwen 2.5 Coder 32B | 21.4 GB | Q2_K |
| Qwen 3 32B | 21.4 GB | Q2_K |
| DeepSeek R1 Distill Qwen 32B | 21.4 GB | Q2_K |
| Gemma 4 31B | 19.8 GB | Q2_K |
| Qwen 3 30B-A3B | 18.8 GB | Q2_K |
| Qwen3-Coder 30B-A3B | 18.8 GB | Q2_K |
| Gemma 3 27B | 17.6 GB | Q3_K_M |
Image and video models
Best weight precision that fits 16 GB with model swapping.
| Model | Best precision |
|---|---|
| Stable Diffusion 1.5 | FP16 / BF16 |
| SDXL 1.0 | FP16 / BF16 |
| Stable Diffusion 3.5 Medium | FP16 / BF16 |
| Stable Diffusion 3.5 Large | FP8 |
| FLUX.1 [dev] | FP8 |
| FLUX.1 [schnell] | FP8 |
| Qwen-Image | GGUF Q5_K_S |
| Wan 2.2 A14B (T2V / I2V) | GGUF Q5_K_S |
| Wan 2.2 TI2V 5B | FP16 / BF16 |
| HunyuanVideo | GGUF Q5_K_S |
| FLUX.1 Kontext [dev] | FP8 |
| FLUX.2 [dev] | Needs offloading or more VRAM |
| Z-Image Turbo | FP16 / BF16 |
| Qwen-Image-Edit 2509 | GGUF Q5_K_S |
| LTX-2 | GGUF Q5_K_S |
GPUs with 16 GB
- RTX 4060 Ti 16GB
- RTX 4070 Ti Super 16GB
- RTX 4080 16GB
- RTX 4080 Super 16GB
- RTX 5060 Ti 16GB
- RTX 5070 Ti 16GB
- RTX 5080 16GB
- Radeon RX 6800 XT 16GB
- Radeon RX 7800 XT 16GB
- Radeon RX 9060 XT 16GB
- Radeon RX 9070 XT 16GB
- Intel Arc A770 16GB
- Tesla T4 16GB
- Apple M4 16GB
- Apple M4 Pro 24GB
16 GB VRAM: common questions
Is 16 GB of VRAM enough for local LLMs?
Yes for models up to about 25.2B at Q4_K_M with 8K context. 16 GB fits 28 of the 58 models tracked here, including Gemma 4 26B-A4B, Mistral Small 3.1 24B, gpt-oss-20b.
What is the biggest model 16 GB can run?
Gemma 4 26B-A4B (25.2B, MoE) at Q4_K_M with 8K context, using about 15.6 GB. Qwen 3.5 35B-A3B also fits at Q2_K, with some quality loss.
Is 16 GB enough for Flux and SDXL?
SDXL: FP16 / BF16. FLUX.1 dev: FP8 (best precision that fits with model swapping).