Best overallGood fit
Qwen2.5 Coder 7B Instruct
The practical local copilot. Supports fill-in-the-middle, so it works as an inline completion model rather than only a chat assistant.
- Params
- 7.6B
- Quant
- Q6_K
- Memory
- 6.8 GB
Reference
Best models for GeForce RTX 3060 12 GB
The cheapest sensible entry into local AI. 12 GB of VRAM matters more than raw speed.
26 models fit11 GB usableEstimated
Top picks
Best on GeForce RTX 3060 12 GB
Runner-upGood fit
Gemma 2 9B Instruct
Unusually good at natural, well-structured prose for its size. The short context is the catch.
- Params
- 9.2B
- Quant
- Q4_K_M
- Memory
- 8.1 GB
ThirdGood fit
Qwen2.5 7B Instruct
The default starting point for local work on 8–12 GB cards. Strong instruction following and reliable tool-call formatting for its size.
- Params
- 7.6B
- Quant
- Q6_K
- Memory
- 6.8 GB
Full results
26 models
Estimated
| # | Model | Params | Licence | Quantization | Memory | ~tok/s | Fit | Fine-tune |
|---|---|---|---|---|---|---|---|---|
| 1 | Qwen2.5 Coder 7B Instruct | 7.6B | Apache 2.0 | Q6_K | 6.8 GB | 45 | Good fit | Yes |
| 2 | Gemma 2 9B Instruct | 9.2B | Gemma Terms of Use | Q4_K_M | 8.1 GB | 50 | Good fit | Yes |
| 3 | Qwen2.5 7B Instruct | 7.6B | Apache 2.0 | Q6_K | 6.8 GB | 45 | Good fit | Yes |
| 4 | Llama 3.1 8B Instruct | 8B | Llama 3.1 Community License | Q6_K | 7.7 GB | 42 | Good fit | Yes |
| 5 | Mistral 7B Instruct v0.3 | 7.25B | Apache 2.0 | Q4_K_M | 5.7 GB | 64 | Excellent fit | Yes |
| 6 | Qwen3 8B | 8.2B | Apache 2.0 | Q5_K_M | 7.1 GB | 48 | Good fit | Yes |
| 7 | Code Llama 7B Instruct | 6.7B | Llama 2 Community License | Q4_K_M | 8.3 GB | 69 | Good fit | Yes |
| 8 | StarCoder2 15B | 16B | BigCode OpenRAIL-M | Q4_K_M | 10 GB | 29 | Tight fit | No |
| 9 | Mistral Nemo 12B Instruct | 12.2B | Apache 2.0 | Q4_K_M | 9.1 GB | 38 | Tight fit | Yes |
| 10 | Qwen3 14B | 14.8B | Apache 2.0 | Q4_K_M | 10 GB | 31 | Tight fit | No |
| 11 | Phi-3.5 Mini Instruct | 3.8B | MIT | Q4_K_M | 5.7 GB | 121 | Excellent fit | Yes |
| 12 | Gemma 3 4B Instruct | 4.3B | Gemma Terms of Use | Q5_K_M | 4.7 GB | 91 | Excellent fit | Yes |
| 13 | Gemma 3 12B Instruct | 12.2B | Gemma Terms of Use | Q4_K_M | 10 GB | 38 | Tight fit | Yes |
| 14 | Llama 3.2 3B Instruct | 3.2B | Llama 3.2 Community License | Q8_0 | 4.6 GB | 82 | Excellent fit | Yes |
| 15 | Qwen2.5 Coder 32B Instruct | 32.8B | Apache 2.0 | Q4_K_M | 21 GB | 14 | Runs with CPU offload | No |
| 16 | Qwen2.5 32B Instruct | 32.8B | Apache 2.0 | Q4_K_M | 21 GB | 14 | Runs with CPU offload | No |
| 17 | SmolLM2 1.7B Instruct | 1.7B | Apache 2.0 | F16 | 5.2 GB | 82 | Excellent fit | Yes |
| 18 | DeepSeek-R1-Distill-Qwen-32B | 32.8B | MIT | Q4_K_M | 21 GB | 14 | Runs with CPU offload | No |
| 19 | Llama 3.2 1B Instruct | 1.24B | Llama 3.2 Community License | F16 | 3.0 GB | 112 | Excellent fit | Yes |
| 20 | Gemma 3 27B Instruct | 27.4B | Gemma Terms of Use | Q4_K_M | 21 GB | 17 | Runs with CPU offload | No |
| 21 | Phi-4 14B | 14.7B | MIT | Q4_K_M | 11 GB | 31 | Runs with CPU offload | No |
| 22 | Mistral Small 24B Instruct | 23.6B | Apache 2.0 | Q4_K_M | 16 GB | 20 | Runs with CPU offload | No |
| 23 | Qwen2.5 14B Instruct | 14.8B | Apache 2.0 | Q4_K_M | 11 GB | 31 | Runs with CPU offload | No |
| 24 | Qwen3 30B-A3B | 30.5B | Apache 2.0 | Q4_K_M | 19 GB | 140 | Runs with CPU offload | No |
| 25 | DeepSeek-R1-Distill-Qwen-14B | 14.8B | MIT | Q4_K_M | 11 GB | 31 | Runs with CPU offload | No |
| 26 | DeepSeek-Coder-V2-Lite Instruct | 15.7B | DeepSeek License | Q4_K_M | 11 GB | 192 | Runs with CPU offload | Yes |
How this page is produced
Memory is calculated from each model’s published geometry against this machine’s usable capacity. Where the vendor publishes a memory bandwidth figure, an order-of-magnitude token rate is derived from it and labelled as an estimate.
- · 15 of these can also be fine-tuned on this hardware.
Check your own machine
ModelLM can read your actual GPU, VRAM and RAM and size every model against it.
Detect my hardwareRelated