NVIDIA GeForce RTX 3060
NVIDIA GeForce RTX 3060 is a mid-range GPU from NVIDIA. It offers 12 GB of VRAM, sufficient for most mid-to-large open-source models. It delivers 12.74 TFLOPS of FP32 compute performance with 360 GB/s of memory bandwidth It can run 20 models in our database. It's the sweet spot for budget-conscious users who still want to run capable local models.
Specifications
| Vendor | NVIDIA |
| Full Name | NVIDIA GeForce RTX 3060 |
| VRAM | 12GB |
| Performance Tier | Mid-Range |
| Benchmark Score | 17,000 |
| FP32 Compute | 12.74TFLOPS |
| Memory Bandwidth | 360GB/s |
| Compatible Models | 20Models that can run on this GPU |
Strengths
- 12 GB VRAM — good for most models up to ~30B parameters
- 360 GB/s bandwidth — good memory throughput
Limitations
- 12.74 TFLOPS — modest compute may limit inference speed
Compatible Models (20)
DeepSeek85%
DeepSeek R1 Distill Qwen 14B Q4_K_M
14.0B
11.9 GB
Q4_K_M
131,072 ctx
Qwen84%
Qwen3 14B Q4_K_M
14.0B
11.9 GB
Q4_K_M
32,768 ctx
Phi83%
Phi-4 14B Q4_K_M
14.0B
11.9 GB
Q4_K_M
16,384 ctx
Llama82%
Llama 3.1 8B Q8_0
8.0B
11.2 GB
Q8_0
8,192 ctx
Gemma82%
Gemma 3 12B Q4_K_M
12.0B
10.2 GB
Q4_K_M
32,768 ctx
DeepSeek82%
DeepSeek R1 Distill Llama 8B Q8_0
8.0B
11.2 GB
Q8_0
131,072 ctx
DeepSeek81%
DeepSeek R1 Distill Qwen 7B Q8_0
7.0B
9.8 GB
Q8_0
131,072 ctx
Yi81%
Yi 1.5 9B Q4_K_M
9.0B
7.7 GB
Q4_K_M
4,096 ctx
Qwen80%
Qwen3 8B Q8_0
8.0B
11.2 GB
Q8_0
32,768 ctx
Llama80%
Llama 3.1 8B Q4_K_M
8.0B
6.8 GB
Q4_K_M
8,192 ctx
Llama80%
Llama 3.1 8B 128K Q4_K_M
8.0B
7.4 GB
Q4_K_M
131,072 ctx
DeepSeek80%
DeepSeek R1 Distill Llama 8B Q4_K_M
8.0B
6.8 GB
Q4_K_M
131,072 ctx
+ 8. Check the Model Library for the complete list.