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

Can I run Nanbeige4.3-33B on an RTX 3090?

Yes, with a GGUF version

The published weights are too big, but the IQ4_XS file from AtomicChat/Nanbeige4.3-33B-GGUF (20.3 GB) fits, with room for ≈ 6K tokens of context.

Full analysis of Nanbeige4.3-33B Everything that runs on RTX 3090 Get the IQ4_XS file on Hugging Face
ollama run hf.co/AtomicChat/Nanbeige4.3-33B-GGUF:IQ4_XS

Files and context on RTX 3090

Each file of Nanbeige4.3-33B (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 3090 Context (16-bit)Context (8-bit)
As published66.2 GBDoesn't fit——
Q8_0AtomicChat/Nanbeige4.3-33B-GGUF35.2 GBDoesn't fit——
Q6_KAtomicChat/Nanbeige4.3-33B-GGUF30.1 GBDoesn't fit——
Q5_KAtomicChat/Nanbeige4.3-33B-GGUF27.9 GBDoesn't fit——
Q5_K_MAtomicChat/Nanbeige4.3-33B-GGUF24.7 GBDoesn't fit——
Nanbeige4.3-33B-Q5_K_MAtomicChat/Nanbeige4.3-33B-GGUF23.5 GBDoesn't fit——
Nanbeige4.3-33B-AD-Q5_K_M-Q4_K_MAtomicChat/Nanbeige4.3-33B-GGUF22.6 GBDoesn't fit——
Q4_K_MAtomicChat/Nanbeige4.3-33B-GGUF21.2 GBDoesn't fit2K4K
IQ4_XSAtomicChat/Nanbeige4.3-33B-GGUF20.3 GBTight6K11K
IQ3_SAtomicChat/Nanbeige4.3-33B-GGUF17.1 GBFits21K42K
IQ3_XXSAtomicChat/Nanbeige4.3-33B-GGUF14.9 GBFits31K61K
IQ2_SAtomicChat/Nanbeige4.3-33B-GGUF13.7 GBFits36K72K

Nanbeige4.3-33B on other hardware

Popular models that run on RTX 3090

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.