NVIDIA GeForce RTX 4080 — Benchmarks & Specs

Bottom line: how fast is the GeForce RTX 4080?

SpecPicks median: 102 tokens/sec across 9 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 59.0–143 tok/s), from knightli.com, llama.cpp GitHub, GPU-Benchmarks-on-LLM-Inference, Hardware Corner (3 runs), MyAIHardware, kunalganglani.com Local LLM Benchmarks, LocalScore.ai. At the $1,501.95 Amazon price of the cheapest linked listing (checked 2026-10-20 16:14 UTC; prices change), that is a SpecPicks Value Index of 67.9 tok/s per $1,000. At 1440p Ultra, the GeForce RTX 4080 averages 118 fps in Tom's Hardware GPU Hierarchy 2026: rasterization geomean (11 games), per Tom's Hardware. For local LLM inference it generates 70.1 tokens/sec running ministral-3:14b at q4_K_M under ollama, per glukhov.org. In 3DMark Fire Strike it scores 48,524 points, per Overclocking.com. Its 16 GB of VRAM is the binding constraint for local inference: the largest model on file at 4-bit quantization on this card is gemma2:27b (26B parameters, IQ4_XS), generating 121.7 tokens/sec under llama.cpp, per glukhov.org. That run used a 19,000-token context; the Will-It-Run matrix sizes each model at its listed quantization with an ~8K context, and by that measure gpt-oss-20b at MXFP4 is the largest that fits in 16 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.