Best LLMs for 8GB VRAM
With 8 GB of VRAM you can run models up to Qwen 3.5 9B (8.95B) at Q4_K_M with an 8K context. 18 of 58 tracked models fit fully, and 7 more fit at a lower quantization.
Models that fit in 8 GB
| Model | VRAM at 8K | Best quant that fits | Max context at Q4 |
|---|---|---|---|
| Qwen 3.5 9B | 6.4 GB | Q5_K_M | 44K |
| Ministral 3 8B | 6.7 GB | Q5_K_M | 14K |
| Qwen 3 8B | 6.6 GB | Q5_K_M | 14K |
| Llama 3.1 8B | 6.3 GB | Q5_K_M | 17K |
| DeepSeek R1 Distill Llama 8B | 6.3 GB | Q5_K_M | 17K |
| Gemma 4 E4B | 5.7 GB | Q6_K | 124K |
| Qwen 2.5 7B | 5.6 GB | Q6_K | 32K |
| DeepSeek R1 Distill Qwen 7B | 5.6 GB | Q6_K | 43K |
| Mistral 7B v0.3 | 5.7 GB | Q6_K | 22K |
| Gemma 4 E2B | 4.0 GB | Q8_0 | 128K |
| Gemma 3 4B | 3.7 GB | Q8_0 | 128K |
| Qwen 3.5 4B | 3.7 GB | Q8_0 | 130K |
| Qwen 3 4B | 4.2 GB | Q8_0 | 31K |
| Phi-4 Mini | 4.1 GB | Q8_0 | 35K |
| Ministral 3 3B | 3.5 GB | Q8_0 | 47K |
| Llama 3.2 3B | 3.5 GB | Q8_0 | 44K |
| 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 8 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 14B | 10.5 GB | Q2_K |
| Qwen 2.5 14B | 10.7 GB | Q2_K |
| Phi-4 14B | 10.6 GB | Q2_K |
| DeepSeek R1 Distill Qwen 14B | 10.7 GB | Q2_K |
| Ministral 3 14B | 9.7 GB | Q2_K |
| Mistral Nemo 12B | 9.0 GB | Q2_K |
| Gemma 3 12B | 8.8 GB | Q3_K_M |
Image and video models
Best weight precision that fits 8 GB with model swapping.
| Model | Best precision |
|---|---|
| Stable Diffusion 1.5 | FP16 / BF16 |
| SDXL 1.0 | FP16 / BF16 |
| Stable Diffusion 3.5 Medium | FP8 |
| Stable Diffusion 3.5 Large | GGUF Q6_K |
| FLUX.1 [dev] | GGUF Q4_K_S / NF4 |
| FLUX.1 [schnell] | GGUF Q4_K_S / NF4 |
| Qwen-Image | Needs offloading or more VRAM |
| Wan 2.2 A14B (T2V / I2V) | Needs offloading or more VRAM |
| Wan 2.2 TI2V 5B | GGUF Q6_K |
| HunyuanVideo | Needs offloading or more VRAM |
| FLUX.1 Kontext [dev] | GGUF Q4_K_S / NF4 |
| FLUX.2 [dev] | Needs offloading or more VRAM |
| Z-Image Turbo | FP8 |
| Qwen-Image-Edit 2509 | Needs offloading or more VRAM |
| LTX-2 | Needs offloading or more VRAM |
GPUs with 8 GB
- RTX 3050 8GB
- RTX 3060 Ti 8GB
- RTX 3070 8GB
- RTX 3070 Ti 8GB
- RTX 4060 8GB
- RTX 4060 Ti 8GB
- RTX 5060 8GB
- RTX 5060 Ti 8GB
- Radeon RX 7600 8GB
8 GB VRAM: common questions
Is 8 GB of VRAM enough for local LLMs?
Yes for models up to about 8.95B at Q4_K_M with 8K context. 8 GB fits 18 of the 58 models tracked here, including Qwen 3.5 9B, Ministral 3 8B, Qwen 3 8B.
What is the biggest model 8 GB can run?
Qwen 3.5 9B (8.95B) at Q4_K_M with 8K context, using about 6.4 GB. Qwen 3 14B also fits at Q2_K, with some quality loss.
Is 8 GB enough for Flux and SDXL?
SDXL: FP16 / BF16. FLUX.1 dev: GGUF Q4_K_S / NF4 (best precision that fits with model swapping).