Model Anatomy See what's inside open-weight AI models

Can I run Mellum2.1-12B-A2.5B-Thinking on an RTX 3060 12 GB?

Yes, with a GGUF version

The published weights are too big, but the Mellum2.1-12B-A2.5B-Thinking-MTP-Q4_K_M file from JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF (8.33 GB) fits, with room for ≈ 128K tokens of context.

Full analysis of Mellum2.1-12B-A2.5B-Thinking Everything that runs on RTX 3060 12 GB Get the Mellum2.1-12B-A2.5B-Thinking-MTP-Q4_K_M file on Hugging Face
ollama run hf.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF

Files and context on RTX 3060 12 GB

Each file of Mellum2.1-12B-A2.5B-Thinking (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.

FileSizeOn RTX 3060 12 GB Context (16-bit)Context (8-bit)
As published24.3 GBDoesn't fit——
Mellum2.1-12B-A2.5B-Thinking-MTP-BF16JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF25.2 GBDoesn't fit——
BF16JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF24.3 GBDoesn't fit——
Mellum2.1-12B-A2.5B-Thinking-MTP-Q8_0JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF13.4 GBDoesn't fit——
Q8_0JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF12.9 GBDoesn't fit——
Mellum2.1-12B-A2.5B-Thinking-MTP-Q6_KJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF11.3 GBDoesn't fit——
Q6_KJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF10.9 GBDoesn't fit——
Mellum2.1-12B-A2.5B-Thinking-MTP-Q4_K_MJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF8.33 GBFits128K128K
Q4_K_MJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF8.07 GBFits128K128K
Mellum2.1-12B-A2.5B-Thinking-MTP-MXFP4_MOEJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF7.27 GBFits128K128K
MXFP4_MOEJetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF7.03 GBFits128K128K

Mellum2.1-12B-A2.5B-Thinking on other hardware

Popular models that run on RTX 3060 12 GB

Assumes 90% of the card's memory is usable and at least 4K tokens of context for one conversation, with the context cache in 16-bit unless noted. Fits leaves at least 20% of that memory for context; Tight leaves less.