NVIDIA GeForce RTX 4070 Ti Super 16GB
$799
The RTX 4070 Ti Super offers excellent mid-range performance for local LLMs and image generation. Great throughput for 7B-13B models with fast CUDA cores.
Specifications
Memory 16GB GDDR6X
Memory Bandwidth 672 GB/s
GPU Cores 8,448
CPU Cores N/A
TDP 285W
Max Model (Q4) 28B parameters
Max Model (Q8) 14B parameters
Performance Tier Mid
Category NVIDIA GPU
Performance Benchmarks
Llama 8B Q4 (tok/s) 85
SDXL 1024px (seconds) 5s
Flux 1024px (seconds) 16s
Pros
- Excellent performance per watt
- Fast image generation (SDXL, FLUX)
- Good balance of price and speed
Cons
- Still limited to 16GB VRAM
- Cannot run 30B+ models
- Requires adequate PSU and cooling
Compatible Models (Q4)
Models that fit in 16GB at Q4 quantization
Llama 3.2 8B Instruct 8B
6GB required Llama 3.2 3B Instruct 3B
3.5GB required Mistral 7B Instruct v0.3 7B
5.5GB required Gemma 3 27B Instruct 27B
15.5GB required Gemma 3 12B Instruct 12B
8GB required DeepSeek Coder V2 Instruct 236B
12.5GB required Qwen2.5 Coder 7B Instruct 7B
5.5GB required StarCoder2 15B 15B
9.5GB required DeepSeek Coder 6.7B Instruct 6.7B
5.35GB required DeepSeek R1 Distill Qwen 7B 7B
5.5GB required Phi-4 14B
9GB required FLUX.1 Schnell 12B
6GB required FLUX.1 Dev 12B
6GB required Stable Diffusion XL Base 1.0 3.5B
3GB required Stable Diffusion 3.5 Large 8B
5GB required SDXL Turbo 3.5B
3GB required HunyuanVideo 8.3B
8GB required LTX-Video 2B
4GB required I2VGen-XL 1.5B
4GB required Kokoro 82M 0.082B
0.5GB required + 3 more models
Compatible at Q8
20 models can run at Q8 quantization