NVIDIA GeForce RTX 4070 — Benchmarks & Specs

Bottom line: how fast is the GeForce RTX 4070?

SpecPicks median: 52.1 tokens/sec across 7 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 38.1–92.3 tok/s), from knightli.com llama.cpp GPU Benchmark Scoreboard, Local LLM Benchmarks, LocalScore.ai, Hardware Corner (4 runs). At the $849.00 Amazon price of the cheapest linked listing (checked 2026-10-20 20:20 UTC; prices change), that is a SpecPicks Value Index of 61.4 tok/s per $1,000. One of those runs: llama3.1:8b at q4_K_M generates 76.3 tokens/sec under llama.cpp, per LocalScore.ai. At 1440p Ultra, the GeForce RTX 4070 averages 91 fps in Cyberpunk 2077, per Overclocking.com. In 3DMark Fire Strike it scores 37,790 points, per Overclocking.com. 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 (14B parameters, q4_K_M), generating 37.9 tokens/sec under llama.cpp, per LocalScore.ai.

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.