NVIDIA GeForce RTX 4070 SUPER — Benchmarks & Specs

Bottom line: how fast is the GeForce RTX 4070 SUPER?

SpecPicks median: 65.6 tokens/sec across 8 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 42.2–109 tok/s), from llama.cpp GitHub Discussions, MyAIHardware, Hardware Corner (3 runs), Mustafa.net LLM Benchmark, LocalScore (2 runs). At the $969.00 Amazon price of the cheapest linked listing (checked 2026-10-21 10:15 UTC; prices change), that is a SpecPicks Value Index of 67.7 tok/s per $1,000. At 1440p Ultra, the GeForce RTX 4070 SUPER averages 112 fps in Elden Ring, per Reviewed. For local LLM inference it generates 40.0 tokens/sec running Generic 13B LLM at q4_K_M under llama.cpp, per Mustafa.net LLM Benchmark. In 3DMark Fire Strike it scores 40,043–40,528 points, across two published results (Overclocking.com, 140429 hardware review). 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 Instruct (14.8B parameters, q4_K_M), generating 24.9 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.