Can I run Qwen2.5-32B-Instruct on an RTX 3060 12 GB?
Yes, with a GGUF version
The published weights are too big, but the IQ2_XXS file from bartowski/Qwen2.5-32B-Instruct-GGUF (9.03 GB) fits, with room for ≈ 7K tokens of context. That file is under 3 bits per weight, so expect noticeably lower quality.
Full analysis of Qwen2.5-32B-Instruct
Everything that runs on RTX 3060 12 GB
Get the IQ2_XXS file on Hugging Face
ollama run hf.co/bartowski/Qwen2.5-32B-Instruct-GGUF:IQ2_XXSFiles and context on RTX 3060 12 GB
Each file of Qwen2.5-32B-Instruct (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.
| File | Size | On RTX 3060 12 GB | Context (16-bit) | Context (8-bit) |
|---|---|---|---|---|
| As published | 65.5 GB | Doesn't fit | — | — |
| F16 | 65.5 GB | Doesn't fit | — | — |
| Q8_0 | 34.8 GB | Doesn't fit | — | — |
| Q6_K_L | 27.3 GB | Doesn't fit | — | — |
| Q6_K | 26.9 GB | Doesn't fit | — | — |
| Q5_K_L | 23.7 GB | Doesn't fit | — | — |
| Q5_K_M | 23.3 GB | Doesn't fit | — | — |
| Q5_K_S | 22.6 GB | Doesn't fit | — | — |
| Q4_K_L | 20.4 GB | Doesn't fit | — | — |
| Q4_K_M | 19.9 GB | Doesn't fit | — | — |
| Q4_K_S | 18.8 GB | Doesn't fit | — | — |
| Q4_0 | 18.7 GB | Doesn't fit | — | — |
| Qwen2.5-32B-Instruct-Q4_0_4_4 | 18.6 GB | Doesn't fit | — | — |
| Qwen2.5-32B-Instruct-Q4_0_4_8 | 18.6 GB | Doesn't fit | — | — |
| Qwen2.5-32B-Instruct-Q4_0_8_8 | 18.6 GB | Doesn't fit | — | — |
| Q3_K_XL | 17.9 GB | Doesn't fit | — | — |
| IQ4_XS | 17.7 GB | Doesn't fit | — | — |
| Q3_K_L | 17.2 GB | Doesn't fit | — | — |
| Q3_K_M | 15.9 GB | Doesn't fit | — | — |
| IQ3_M | 14.8 GB | Doesn't fit | — | — |
| Q3_K_S | 14.4 GB | Doesn't fit | — | — |
| IQ3_XS | 13.7 GB | Doesn't fit | — | — |
| Q2_K_L | 13.1 GB | Doesn't fit | — | — |
| Q2_K | 12.3 GB | Doesn't fit | — | — |
| IQ2_M | 11.3 GB | Doesn't fit | — | — |
| IQ2_S | 10.4 GB | Doesn't fit | 2K | 3K |
| IQ2_XS | 9.96 GB | Doesn't fit | 3K | 6K |
| IQ2_XXS | 9.03 GB | Tight | 7K | 13K |