NVIDIA
RTX 3070 Ti
8 Go
VRAM
—
Best generation
—
Efficiency (ref.)
—
Elec. cost / M tokens (ref.) · 0,25 €/kWh
What this card can run
Gemma 3 270M Q4_K_M
0.25 Go
Qwen2.5 0.5B Q4_K_M
0.4 Go
Qwen2.5 Coder 0.5B Q4_K_M
0.4 Go
Qwen3 0.6B Q4_K_M
0.48 Go
Qwen3.5 0.8B Q4_K_M
0.58 Go
Llama 3.2 1B Q4_K_M
0.81 Go
Qwen2.5 1.5B Q4_K_M
0.99 Go
Qwen2.5 Coder 1.5B Q4_K_M
0.99 Go
R1 Distill 1.5B Qwen Q4_K_M
1.12 Go
Qwen3 1.7B Q4_K_M
1.28 Go
Qwen3.5 2B Q4_K_M
1.4 Go
Gemma 2 2B Q4_K_M
1.71 Go
Qwen2.5 3B Q4_K_M
1.93 Go
Qwen2.5 Coder 3B Q4_K_M
1.93 Go
Llama 3.2 3B Instruct Q4_K_M
2.02 Go
Qwen3 2507 4B Q4_K_M
2.5 Go
Qwen3 4B Q4_K_M
2.5 Go
Gemma 3n E2B Q4_K_M
2.79 Go
Qwen3.5 4B Q4_K_M
3.01 Go
Gemma 4 E2B Q4_K_M
3.46 Go
Gemma 3n E4B Q4_K_M
4.24 Go
Qwen2.5 7B Instruct Q4_K_M
4.68 Go
Qwen2.5 Coder 7B Q4_K_M
4.68 Go
R1 Distill 7B Qwen Q4_K_M
4.68 Go
Meta Llama 3.1 8B Instruct Q4_K_M
4.92 Go
Qwen3 8B Q4_K_M
5.03 Go
R1 0528 (Qwen3) 8B Q4_K_M
5.03 Go
Gemma 4 E4B Q4_K_M
5.41 Go
Gemma 2 9B Q4_K_M
5.76 Go
Qwen3.5 9B Q4_K_M
6.17 Go
Gemma 4 12B Q4_K_M
7.66 Go
Qwen2.5 14B Instruct Q4_K_M
8.99 Go
DeepSeek R1 Distill Qwen 14B Q4_K_M
8.99 Go
Qwen2.5 Coder 14B Q4_K_M
8.99 Go
Qwen3 14B Q4_K_M
9 Go
Phi-4 Q4_K_M
9.05 Go
Coder V2 Lite 16B-A2.4B Q4_K_M
10.36 Go
Smooth ≥ 60 tok/s (faster than you read) · Comfortable 30-60 · Slow < 30 · "Doesn't fit" = model size + 2 GB headroom > VRAM (CPU offloading is excluded from the protocol: the numbers would be meaningless).
Measurements by model
No runs for this card yet. Your move: run the bench.
VRAM used = peak system GPU memory during the run (includes the OS, roughly 0.5–1 GB more than the model alone).