What can a RTX 3060 12GB run?
12GB · 360 GB/s · 170W · $220
The budget hero: 12 GB and solid CUDA support make it everyone's first local-LLM card.
Runs great14
Comfortable headroom — daily-driver territory.
| Model | Quant | Est. speed | Max context |
|---|---|---|---|
| deepseek moe 16b chat | FP16 | Infinity tok/s | — |
| qwen2 7b instruct | FP16 | Infinity tok/s | — |
| Qwen3-0.6B | FP16 | 180 tok/s | 33k |
| Gemma-3n-E4B | Q8_0 | 150 tok/s | 70k |
| SmolLM2-1.7B | FP16 | 66 tok/s | 8k |
| Phi-4-mini | Q8_0 | 55 tok/s | 71k |
| Gemma-3-4B | Q8_0 | 54 tok/s | 71k |
| Qwen3-4B | Q8_0 | 54 tok/s | 33k |
| Mistral-7B-Instruct | Q6_K | 37 tok/s | 33k |
| DeepSeek-R1-Distill-Qwen-7B | Q6_K | 36 tok/s | 37k |
| Qwen2.5-Coder-7B | Q6_K | 36 tok/s | 37k |
| Llama-3.2-3B | FP16 | 35 tok/s | 47k |
| Llama-3.1-8B | Q6_K | 34 tok/s | 34k |
| Qwen3-8B | Q6_K | 34 tok/s | 33k |
Tight squeeze11
It fits, but close the browser tabs and mind the context.
| Model | Quant | Est. speed | Max context |
|---|---|---|---|
| GPT-OSS-20B | IQ2_M | 183 tok/s | 23k |
| Codestral-22B | IQ2_M | 30 tok/s | 20k |
| Mistral-Small-24B | IQ2_M | 27 tok/s | 15k |
| Phi-4-14B | Q4_K_M | 25 tok/s | 11k |
| DeepSeek-R1-Distill-Qwen-14B | Q4_K_M | 25 tok/s | 11k |
| Qwen3.8 | IQ2_M | 25 tok/s | 8k |
| Qwen3.8 27B | IQ2_M | 25 tok/s | 8k |
| Gemma-3-27B | IQ2_M | 24 tok/s | 8k |
| Gemma-3-12B | Q5_K_M | 24 tok/s | 9k |
| qwen35 27B | IQ2_M | 24 tok/s | 8k |
| Qwen3-14B | MLX 4-bit | 23 tok/s | 7k |
Measured on this card0
Everything here was benchmarked on this exact card by someone who owns one — the tables above are our arithmetic; these numbers came off real machines.
Nobody has measured anything on this card yet. Paste a benchmark — it takes one keystroke and no form.