NVIDIA GeForce RTX 4060 Ti 16GB, AI & Machine Learning Benchmarks & Specs

16GB · AI Score 3.2/100 · first-party measured on 12 AI workloads

3.2 AI Score Includes estimates

Every number on this page is first-party: NVIDIA GeForce RTX 4060 Ti was run on our pinned 12-workload AI suite on 2026-07-11, with under 0.5% run-to-run variance. On Llama 3.1 8B (Q4_K_M) NVIDIA GeForce RTX 4060 Ti delivers about 57.09 tokens/sec. Llama 3.3 70B does not fit. It needs roughly 42GB and this card has 16GB. For image generation, SDXL runs at 2.29 it/s, while FLUX.1-dev won't fit at BF16 (needs ~26GB). 6 of the 12 workloads won't fit on 16GB at the tested precision, Qwen3 32B, Llama 3.3 70B, FLUX.1-dev, FLUX.1 Kontext and others. We publish those as hard gates rather than quietly dropping to a smaller quant.

Bench notes: from the person who ran it

Important: this is the 16GB SKU, I say that everywhere because the 8GB version is a different card for AI. The 16GB ran LTX video and Z-Image, stuff even 12GB cards refuse. It did creep to 103% of its 160W rating on SDXL and hit 77°C on video, but for a budget AI card the VRAM does the talking. Quick note on the setup: all my AI benchmarking was done on rented cloud GPUs, I used all three of Vast.ai, RunPod and Modal depending on which had the card, and they all have their pros and cons. Same pinned harness on every run, and everything here got double-checked before it went up.

AI & Machine Learning benchmark results

Text Generation tok/s 30

Qwen3 0.6B399.21

Llama 3.2 1B283.32

MiniCPM5 2B158.1

LFM2.5 2.6B142.22

Granite 4.1 3B108.51

gpt-oss-20b102.24

Qwen3 4B96.29

Spark-X2.5-4B91.93

Nemotron 3 Nano 4B88.1

Agents-A1-4B82.98

DeepSeek Coder 7B Instruct v1.562.54

Qwen2-7B-Instruct58.84

Workload	Result	Telemetry	Data
Qwen3 0.6B	399.21 tok/s	54 W55°CQ4_K_M	✓ Measured
Llama 3.2 1B	283.32 tok/s	69 W57°CQ4_K_M	✓ Measured
MiniCPM5 2B	158.1 tok/s	73 W50°CQ4_K_M	✓ Measured
LFM2.5 2.6B	142.22 tok/s	80 W50°CQ4_K_M	✓ Measured
Granite 4.1 3B	108.51 tok/s	81 W57°CQ4_K_M	✓ Measured
Agents-A1-4B	82.98 tok/s	101 W56°CQ4_K_M	✓ Measured
Nemotron 3 Nano 4B	88.1 tok/s	86 W46°CQ4_K_M	✓ Measured
Qwen3 4B	96.29 tok/s	2.8 GB peak99 W52°C0.97 tok/WQ4_K_M	✓ Measured
Spark-X2.5-4B	91.93 tok/s	97 W49°CQ4_K_M	✓ Measured
DeepSeek Coder 7B Instruct v1.5	62.54 tok/s	93 W54°CQ4_K_M	✓ Measured
OLMo 3 7B Instruct	58.31 tok/s	104 W56°CQ4_K_M	✓ Measured
OLMo 3 7B Think	58.29 tok/s	107 W56°CQ4_K_M	✓ Measured
Qwen2-7B-Instruct	58.84 tok/s	109 W54°CQ4_K_M	✓ Measured
Qwen2.5-7B	58.47 tok/s	107 W62°CQ4_K_M	✓ Measured
Qwen2.5-Coder 7B	58.48 tok/s	107 W60°CQ4_K_M	✓ Measured
Apertus-8B-Instruct	53.92 tok/s	110 W57°CQ4_K_M	✓ Measured
Llama 3 8B	55.37 tok/s	93 W53°CQ4_K_M	✓ Measured
Llama 3.1 8B	57.09 tok/s	4.8 GB peak111 W55°C0.51 tok/WQ4_K_M	✓ Measured
Qwen3 8B	54.34 tok/s	102 W56°CQ4_K_M	✓ Measured
Nemotron Nano 9B v2	39.24 tok/s	90 W53°CQ4_K_M	✓ Measured
Ornith 1.5 9B	48.42 tok/s	100 W52°CQ4_K_M	✓ Measured
Gemma 4 12B	35.25 tok/s	104 W58°CQ4_K_M	✓ Measured
Qwen2.5-Coder 14B	31.12 tok/s	8.5 GB peak126 W62°C0.25 tok/WQ4_K_M	✓ Measured
Qwen3 14B	30.4 tok/s	110 W59°CQ4_K_M	✓ Measured
gpt-oss-20b	102.24 tok/s	83 W59°CQ4_K_M	✓ Measured
Gemma 4 26B A4B	✕ Won't fit needs ~20 GB	VRAM-gated at this precision	Est.
Qwen3 30B A3B	✕ Won't fit needs ~20 GB	VRAM-gated at this precision	Est.
Gemma 4 31B	✕ Won't fit needs ~22 GB	VRAM-gated at this precision	Est.
Qwen3 32B	✕ Won't fit needs ~23 GB	VRAM-gated at this precision	✓ Measured
Llama 3.3 70B	✕ Won't fit needs ~46 GB	VRAM-gated at this precision	✓ Measured

Image Generation images/min 15

SD Turbo361.33

SDXL Turbo255.5

LCM DreamShaper v794.29

Stable Diffusion 2.128.74

DreamShaper XL Lightning27.95

Stable Diffusion 1.526.23

DreamShaper XL Turbo16.44

Sana 1.6B11.91

PixArt-Sigma XL6.62

Stable Diffusion XL4.58

SSD-1B4.49

Playground v2.52.89

Workload	Result	Telemetry	Data
Stable Diffusion 1.5	26.23 images/min	135 W58°C	✓ Measured
SD Turbo	361.33 images/min	52 W57°C	✓ Measured
Stable Diffusion 2.1	28.74 images/min	135 W54°C	✓ Measured
LCM DreamShaper v7	94.29 images/min	113 W54°C	✓ Measured
SDXL Turbo	255.5 images/min	64 W51°C	✓ Measured
SSD-1B	4.49 images/min	139 W66°C	✓ Measured
Z-Image Turbo	1.41 images/min		✓ Measured 3 hosts ±9%
Sana 1.6B	11.91 images/min	140 W53°C	✓ Measured
Stable Diffusion XL	4.58 images/min	14.5 GB peak161 W75°C13.1 s/img	✓ Measured
DreamShaper XL Lightning	27.95 images/min	132 W62°C	✓ Measured
DreamShaper XL Turbo	16.44 images/min	147 W54°C	✓ Measured
Playground v2.5	2.89 images/min	158 W72°C	✓ Measured
PixArt-Sigma XL	6.62 images/min	140 W57°C	✓ Measured
FLUX.2 klein 4B	✕ Won't fit needs ~19 GB	VRAM-gated at this precision	✓ Measured
FLUX.1 dev	✕ Won't fit needs ~26 GB	VRAM-gated at this precision	✓ Measured

Image Editing images/min 2

Workload	Result	Telemetry	Data
FLUX.1 Kontext dev	✕ Won't fit needs ~26 GB	VRAM-gated at this precision	✓ Measured
Qwen-Image-Edit	✕ Won't fit needs ~42 GB	VRAM-gated at this precision	✓ Measured

Image to Video clips/min 5

LTX-Video (image to video)0.962

Stable Video Diffusion0.643

Wan 2.2 TI2V-5B (image to video)0.368

Stable Video Diffusion XT0.361

CogVideoX-5B I2V0.099

Workload	Result	Telemetry	Data
Stable Video Diffusion	0.64 clips/min	154 W73°C	✓ Measured
LTX-Video (image to video)	0.96 clips/min		✓ Measured 2 hosts ±7% · CPU offload
Wan 2.2 TI2V-5B (image to video)	0.37 clips/min	127 W71°C	✓ Measured 2 hosts ±4% · CPU offload
Stable Video Diffusion XT	0.36 clips/min	147 W74°C	✓ Measured 2 hosts ±0% · CPU offload
CogVideoX-5B I2V	0.1 clips/min	142 W62°C	✓ Measured 2 hosts ±1% · CPU offload

Video Generation frames/s 4

LTX-Video (distilled)1.668

Wan 2.1 1.3B0.176

CogVideoX-2B0.143

Workload	Result	Telemetry	Data
Wan 2.1 1.3B	0.18 frames/s	139 W75°C278.7 s/clip	✓ Measured CPU offload
CogVideoX-2B	0.14 frames/s	145 W62°C343.3 s/clip	✓ Measured 2 hosts ±0% · CPU offload
LTX-Video (distilled)	1.67 frames/s	114 W65°C58.1 s/clip	✓ Measured 3 hosts ±5% · CPU offload
Wan 2.2 5B (720p)	✕ Won't fit needs ~18 GB	VRAM-gated at this precision	✓ Measured

How we measured this. Every result comes from our own pinned, reproducible AI suite, 12 workloads: the Qwen3-4B to Llama-70B LLM ladder (llama.cpp, Q4_K_M), SDXL / Z-Image / FLUX-dev generation, FLUX-Kontext / Qwen-Edit editing, and LTX / Wan video, run first-party on rented hardware with under 0.5% run-to-run variance. Peak VRAM, power draw, temperature and tokens-per-watt are captured per workload. “Won’t fit” rows are real data: where a model exceeds the card’s VRAM at the tested precision we record a hard gate rather than silently dropping to a smaller quant. Measured 2026-07-11 · harness 2.0.0-standalone.

NVIDIA GeForce RTX 4060 Ti 16GB specifications

Architecture	Ada Lovelace
CUDA cores	4,352
VRAM	16GB GDDR6
Memory bus	128-bit
Memory bandwidth	288 GB/s
Boost clock	2,535 MHz
TDP	165 W
Process	4 nm
Interface	PCIe 4.0 x16
Release date	2023-07-18
Launch MSRP	$499

Verdict, capable, but 16GB sets the ceiling

NVIDIA GeForce RTX 4060 Ti scores 3.2/100, #53 of 102. It ran 6 of 12; 6 exceeded its 16GB. Every figure here is our own measurement.

Relative performance: where the NVIDIA GeForce RTX 4060 Ti 16GB lands

100% = this card, AI & Machine Learning headline metric (AI Score). #18 of 61 desktop cards in this vertical.

GPU	%	AI Score
NVIDIA GeForce RTX 4070 Ti Super	134%	4.3
AMD Radeon RX 7900 XTX	116%	3.7
GeForce RTX 5060 Ti	116%	3.7
NVIDIA GeForce RTX 3080 Ti	109%	3.5
NVIDIA GeForce RTX 4060 Ti 16GB	100%	3.2
NVIDIA GeForce RTX 4070 Ti	100%	3.2
GeForce RTX 5070	100%	3.2
NVIDIA GeForce RTX 4070 Super	97%	3.1
NVIDIA GeForce RTX 4070	97%	3.1

One tier up: NVIDIA GeForce RTX 3080 Ti (+9%) · compare head-to-head
One tier down: NVIDIA GeForce RTX 4070 Ti (0%) · compare head-to-head

Same card, other workloads: NVIDIA GeForce RTX 4060 Ti 16GB Gaming benchmarks

← All AI & Machine Learning GPU rankings

The silicon

Transistors	22,900 million
Die size	187.8 mm²
Process node	4 nm
Fabricated by	TSMC
Transistor density	121.9 million per mm²

Denser than 96% of the 746 cards we have silicon data for. Density is the clearest measure of what a process node bought: a card that gained it without growing the die got its speed from the fab rather than the architecture.

Silicon figures from Wikipedia (CC BY-SA 4.0). Benchmarks on this page are our own. Compare every chip.

What this card can build

Whole-job timings, composed from our measured per-model results on this card.

Workflow	Time	Energy	Basis
Full codebase review	32.1 min	67.32 Wh	measured
Animate a batch of images	54.3 min	115.22 Wh	measured

Can't run: Product photo shoot (needs FLUX.1 Kontext dev), Photo restoration batch (needs FLUX.1 Kontext dev), Restore and enlarge photos (needs FLUX.1 Kontext dev), Character sheet, 12 poses (needs FLUX.1 dev), Short social clips (needs Qwen3 32B), 60-second AI short film (needs Qwen3 32B), 24-frame storyboard (needs Qwen3 32B), 6-panel comic page (needs Qwen3 32B), Long-form article batch (needs Llama 3.3 70B).

Rent or buy?

This card is $499 to buy. The cheapest listed rate on Vast.ai is $0.109/hour, but that is the floor: we budget $0.131/hour, a 20% premium, because idle time, storage and unavailable cheap instances all land on the same bill. At that rate buying wins after 3,815 GPU-hours. Below it you are paying for idle silicon.

How you would use it	GPU-hours a year	Rental cost a year	Time to break even
2 hours a day, hobby	730	$95	5.2 years
8 hours a day, working on it	2,920	$382	1.3 years
24/7, always-on agent	8,760	$1,146	5.2 months

At hobby usage this card is very unlikely to pay for itself before it is superseded. Rent it. Rental figures include a 20% premium over the cheapest listed rate. Ignores electricity, resale and the fact that a rented card can be a newer one tomorrow.

Rental price

$0.109/hr+6.9% since 2026-09-30low $0.092 · high $0.124

Cheapest of the RunPod and Vast on-demand rates we see, sampled daily. Spot and interruptible pricing runs lower.