Written by Jakub Rusinowski · Last updated July 12, 2026
8 GB VRAM with significantly higher bandwidth than the 3070. 58 t/s on 8B models. Available used for $180–240. Best Ampere 8GB card for local AI inference speed.
| VRAM | 8 GB |
| Memory Bandwidth | 608 GB/s |
| TDP | 290 W |
| Architecture | Ampere GA104 |
| Release Year | 2021 |
| MSRP at Launch | $599 |
| Inference Speed (Llama 3.1 8B Q4_K_M) | 42–80 tok/s (estimated) |
| Inference Speed (Llama 3.3 70B Q4_K_M) | Does not fit — needs ~44 GB of 8 GB usable |
All models below run comfortably in 8 GB VRAM with Q4_K_M quantization.
| Llama 3.1 Family | Llama 3.1 8B Instruct · 6 GB VRAM · Q4_K_M · ollama run llama3.1 |
| Llama 3.2 Family | Llama 3.2 11B Vision Instruct · 7 GB VRAM · Q4_K_M · llama-3-2 |
| Qwen 2.5 Family | Qwen 2.5 7B Instruct · 5 GB VRAM · Q4_K_M · ollama run qwen2.5:7b |
| Gemma 2 Family | Gemma 2 9B IT · 6 GB VRAM · Q4_K_M · ollama run gemma2 |
| Phi-4 Mini | Phi-4 Mini (3.8B) · 3 GB VRAM · Q4_K_M · ollama run phi4-mini |
| SmolLM2 | SmolLM2 1.7B Instruct · 2 GB VRAM · Q4_K_M · ollama run smollm2:1.7b |
Install Ollama then run the recommended model for this GPU:
ollama run llama3.1:8b
Yes — the NVIDIA GeForce RTX 3070 Ti has 8 GB VRAM and runs 8 GB VRAM with significantly higher bandwidth than the 3070. 58 t/s on 8B models. Available used for $180–240. Best Ampe
The NVIDIA GeForce RTX 3070 Ti is estimated to run Llama 3.1 8B at 42–80 tok/s with Q4_K_M quantization. Llama 3.3 70B does not fit: it needs about 44 GB against 8 GB usable. These are modelled estimates, not measurements — see /en/methodology.
With 8 GB you can run: Llama 3.1 Family, Llama 3.2 Family, Qwen 2.5 Family, Gemma 2 Family, Phi-4 Mini. Use Ollama for the easiest setup: ollama run llama3.1:8b.
← All GPU Reviews | Check Your Hardware | Full Benchmarks | Can I Run It?