TokenAssemble

Can an NVIDIA GeForce RTX 3060 12 GB run gpt-oss-20b?

Yes, with CPU offload. gpt-oss-20b needs 13.0GB at MXFP4 with llama.cpp and 8K context; the NVIDIA GeForce RTX 3060 12 GB has 12GB of VRAM (11.5GB usable), so part of the model spills to system RAM — it loads, but expect the interactive speed tier.

NVIDIA GeForce RTX 3060 12 GB · gpt-oss-20b · MXFP4

⚠️ Tight — grade C

InteractivebetaHybrid attention — KV computed as full-cache upper bound · MoE speed estimate — wide error bars · CPU offload — speed varies heavily with layer split
weights 12.1 GBKV cache 0.40 GBoverhead 0.44 GBpool 11.5 GBneeds 13.0 GB — spills to system RAM

assumes llama.cpp · batch 1 · 8K context · fp16 KV cache · 64GB system RAM (CPU offload)

$329 MSRP · as of 2026-07-08

Affiliate link — we may earn a commission. Verdicts are computed before any link is attached.

Every quant, graded

NVIDIA GeForce RTX 3060 12 GB × gpt-oss-20b, llama.cpp, 8K context.

QuantFile sizeFitSpeed tier
MXFP412.1 GBTight · Cinteractive · offload · beta

gpt-oss-20b on similar hardware

More models on the NVIDIA GeForce RTX 3060 12 GB

Verdict computed from published specs and measured GGUF file sizes — see how the math works.