NVIDIA GeForce RTX 3060 12 GB — Benchmarks & Specs

Bottom line: how fast is the NVIDIA GeForce RTX 3060 12 GB?

SpecPicks median: 55.2 tokens/sec across 21 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 31.9–75.9 tok/s), from llama.cpp GitHub (2 runs), GitHub: XiongjieDai/GPU-Benchmarks-on-LLM-Inference, TyoLab (7 runs), singhajit.com (6 runs), smeltcore.com (2 runs), GitHub Gist, LocalScore, ModelFit.io. At the $478.00 Amazon price of the cheapest linked listing (checked 2026-10-21 12:20 UTC; prices change), that is a SpecPicks Value Index of 115.5 tok/s per $1,000. At 1440p High with DLSS Quality upscaling, the NVIDIA GeForce RTX 3060 12 GB averages 65 fps in Hogwarts Legacy, per Gamersleaguepros. For local LLM inference it generates 32.8 tokens/sec running llama2:13b at q4_K_M under llama.cpp, per singhajit.com. In Geekbench 6 Vulkan it scores 91,383 points, per Geekbench Browser. Its 12 GB of VRAM is the binding constraint for local inference: the largest model on file at 4-bit quantization on this card is qwen2.5:14b (14.8B parameters, q4_K_M), generating 26.4 tokens/sec under llama.cpp, per LocalScore. That run's source does not report a context length; the Will-It-Run matrix sizes each model at its listed quantization with an ~8K context, and by that measure Phi-4 14B at Q4_K_M is the largest that fits in 12 GB.

Every figure above is a row in the tables below, and each row links out to the review or public benchmark database the number was taken from; the SpecPicks median is computed here from the runs it names, each a row in the AI table, and the per-$1,000 figure divides it by the dated listing price. SpecPicks aggregates published measurements; it does not report first-party benchmark runs.