Can I run it?

Exact model x GPU verdicts: fit, memory math with KV-cache, estimated speed, best quant and upgrade path. Estimates, not measurements.

RTX 3060

12 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MSlow GPT-OSS 20B MXFP4Slow Qwen3.8 27B Q4_K_MNo Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 3060 guide

RTX 4060

8 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MNo GPT-OSS 20B MXFP4No Qwen3.8 27B Q4_K_MNo Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 4060 guide

RTX 4060 Ti

16 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MSlow Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 4060 Ti guide

RTX 4070

12 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MSlow GPT-OSS 20B MXFP4Slow Qwen3.8 27B Q4_K_MNo Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 4070 guide

RTX 4070 Ti SUPER

16 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MSlow Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 4070 Ti SUPER guide

RTX 3090

24 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MRuns Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 3090 guide

RTX 4090

24 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MRuns Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 4090 guide

RTX 5070 Ti

16 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MSlow Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 5070 Ti guide

RTX 5080

16 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MSlow Llama 3.3 70B Instruct Q4_K_MNo

Full RTX 5080 guide

RTX 5090

32 GB

Qwen3 8B Q4_K_MRuns Qwen3 14B Q4_K_MRuns GPT-OSS 20B MXFP4Runs Qwen3.8 27B Q4_K_MRuns Llama 3.3 70B Instruct Q4_K_MSlow

Full RTX 5090 guide