RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

Can PaddleOCR VL 1.6 run on the RTX 4070 Ti Super?

Yes. At 8K context the best fit is F16, using about 2.5 GB of the 15.2 GB usable on the RTX 4070 Ti Super. Estimated speed ceiling: ~351 tokens/sec.

Newer compressed-attention design: long-context figures are an upper bound.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K)1.0 GB Fits1.1 GB Fits1.5 GB Fits3.2 GB Fits1869
3-bit (Q3_K_M)1.1 GB Fits1.2 GB Fits1.6 GB Fits3.3 GB Fits1434
4-bit (Q4_K_M)1.2 GB Fits1.3 GB Fits1.7 GB Fits3.4 GB Fits1147
5-bit (Q5_K_M)1.3 GB Fits1.4 GB Fits1.8 GB Fits3.5 GB Fits984
6-bit (Q6_K)1.4 GB Fits1.5 GB Fits1.9 GB Fits3.6 GB Fits855
8-bit (Q8_0)1.6 GB Fits1.7 GB Fits2.1 GB Fits3.8 GB Fits660
16-bit (F16/BF16)2.5 GB Fits2.5 GB Fits2.9 GB Fits4.6 GB Fits351

• real GGUF file size. Speed is a bandwidth ceiling (672 GB/s), not a benchmark. Methodology.

What fits instead

"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.

More

PaddleOCR VL 1.6 on every GPU and Mac · Every model that runs on the RTX 4070 Ti Super · Other models that fit 16 GB VRAM · Try other settings in the calculator

Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs

Data: Hugging Face Hub (PaddlePaddle/PaddleOCR-VL-1.6), NVIDIA specs. Model data last checked 2026-11-18; refreshed hourly.