NVIDIA GeForce RTX 4070 Ti — Benchmarks & Specs

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

SpecPicks median: 78.6 tokens/sec across 7 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 57.5–97.0 tok/s), from Markaicode, GPU-Benchmarks-on-LLM-Inference (4 runs), Kunal Ganglani LLM Benchmarks, Hardware Corner. At the $1,149.00 Amazon price of the cheapest linked listing (checked 2026-10-13 22:17 UTC; prices change), that is 68.4 tok/s per $1,000. At 1440p Ultra, the GeForce RTX 4070 Ti averages 127 fps in Far Cry 6, per Beebom. For local LLM inference it generates 97.0 tokens/sec running llama3.1:8b at q4_K_M under llama.cpp, per Markaicode. In PassMark G3D Mark it scores 31,573 pts, per PassMark. 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 deepseek-r1:14b (14B parameters, q4_K_M), generating 49.7 tokens/sec under llama.cpp, per geerlingguy/ai-benchmarks.

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.