FR
Catalogue · 26 models

How much VRAM for each LLM?

Memory needed in Q4, Q8 and FP16 with an 8k-token context, and the cheapest machine that runs it. Click a model for details by context and by card.

All26Llama3Qwen8OpenAI2Google2Mistral4Microsoft1DeepSeek3GLM2Kimi1
8 results
Dense · Qwen2025-04
Qwen3 4B4.0 B · dense
4 GBQ46 GBQ810 GBFP16

From: RTX 3060 12 GB · €1,180

Dense · Qwen2025-04
Qwen3 8B8.2 B · dense
7 GBQ411 GBQ819 GBFP16

From: RTX 3060 12 GB · €1,180

Dense · Qwen2025-04
Qwen3 14B15 B · dense
11 GBQ418 GBQ833 GBFP16

From: RTX 3060 12 GB · €1,180

Dense · Qwen2026
Qwen3.8 27B28 B · dense
20 GBQ433 GBQ861 GBFP16

From: 2 × RTX 3060 12 GB · €1,460

MoE · Qwen2025-04
Qwen3 30B-A3B31 B · 3.3 B active
20 GBQ435 GBQ865 GBFP16

From: 2 × RTX 3060 12 GB · €1,460

Dense · Qwen2025-04
Qwen3 32B33 B · dense
23 GBQ439 GBQ871 GBFP16

From: 2 × RTX 3060 12 GB · €1,460

Dense · Qwen2024-11
Qwen2.5 Coder 32B33 B · dense
23 GBQ439 GBQ871 GBFP16

From: 2 × RTX 3060 12 GB · €1,460

MoE · Qwen2025-04
Qwen3 235B-A22B235 B · 22 B active
150 GBQ4263 GBQ8495 GBFP16

From: Mac Studio M3 Ultra 512 GB · €11,500