What LLM can I run? Pick your GPU / VRAM / RAM
Choose your hardware below. Every model that fits is listed with its best quant — computed from measured GGUF file sizes, 8K context, ~8% headroom. Data refreshes daily from Hugging Face.
Best LLMs for a 24GB GPU (RTX 3090 / 4090) — static list
| Model | Best quant that fits | VRAM needed | Params |
|---|---|---|---|
| flint | UD-IQ3_XXS | 18.9 GB | 40B |
| MiniCPM-V-4.7-35B-A3B | IQ3_M | 20.0 GB | 35B |
| granite-4.2-30b | Q4_K_S | 21.6 GB | 30B |
| Xing4.0-29B-A4B | Q4_K_M | 20.8 GB | 29B |
| Qwen3.8-27B | UD-Q4_K_XL | 21.1 GB | 27B |
| Ternary-Bonsai-2-27B | Q2_0 | 10.1 GB | 27B |
| gemma-4-26B-A4B-it | UD-Q4_K_XL | 20.4 GB | 25B |
| mistralai_Mistral-Small-3.2-24B-Instruct-2506 | Q5_K_M | 21.6 GB | 24B |
| LFM2.5-24B-A2B-Turbo-Brilliance-50X-NEO-MAX-Imatrix | IQ4_XS | 16.9 GB | 24B |
| Mistral-Small-3.2-24B-Instruct-2506 | UD-Q5_K_XL | 21.6 GB | 24B |
| Qwen3-14B | Q8_0 | 20.3 GB | 14B |
| Ministral-3-14B-Instruct-2512 | Q8_0 | 18.7 GB | 14B |
Full ranked list: best LLM for 24GB VRAM. Other tiers: 8GB · 12GB · 16GB · 32GB.