Can I run PaddleOCR VL 1.6 locally?
PaddleOCR VL 1.6 by PaddlePaddle has 959M parameters. At 4-bit with an 8K context it needs about 1.3 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.
GGUF Q8_0 file was 49% of the size expected from the parameter count, so this page uses size estimates instead of real file sizes.
This model uses a newer compressed or sparse attention design. Our KV-cache numbers use the standard formula, so treat long-context figures as an upper bound.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 1.0 GB | 1.1 GB | 1.5 GB | 3.2 GB |
| 3-bit (Q3_K_M) | usable | 1.1 GB | 1.2 GB | 1.6 GB | 3.3 GB |
| 4-bit (Q4_K_M) | good, most popular | 1.2 GB | 1.3 GB | 1.7 GB | 3.4 GB |
| 5-bit (Q5_K_M) | very good | 1.3 GB | 1.4 GB | 1.8 GB | 3.5 GB |
| 6-bit (Q6_K) | near lossless | 1.4 GB | 1.5 GB | 1.9 GB | 3.6 GB |
| 8-bit (Q8_0) | effectively lossless | 1.6 GB | 1.7 GB | 2.1 GB | 3.8 GB |
| 16-bit (F16/BF16) | original weights | 2.5 GB | 2.5 GB | 2.9 GB | 4.6 GB |
Which GPUs and Macs run PaddleOCR VL 1.6?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
| Hardware | Memory | Fits? | Best quant | Needs | Max tok/s (est.) | |
|---|---|---|---|---|---|---|
| RTX 3060 12GB | 12 GB | Fits | F16 | 2.5 GB | 188 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | F16 | 2.5 GB | 150 | details |
| RTX 3090 | 24 GB | Fits | F16 | 2.5 GB | 488 | details |
| RTX 4090 | 24 GB | Fits | F16 | 2.5 GB | 526 | details |
| RTX 5090 | 32 GB | Fits | F16 | 2.5 GB | 935 | details |
| RTX 5070 Ti | 16 GB | Fits | F16 | 2.5 GB | 467 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 2.5 GB | 501 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 2.5 GB | 142 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 2.5 GB | 285 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 2.5 GB | 427 | details |
| DGX Spark | 128 GB | Fits | F16 | 2.5 GB | 142 | details |
| RTX 4070 | 12 GB | Fits | F16 | 2.5 GB | 263 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | F16 | 2.5 GB | 234 | details |
| RTX 4060 | 8 GB | Fits | F16 | 2.5 GB | 142 | details |
| Arc B580 | 12 GB | Fits | F16 | 2.5 GB | 238 | details |
| RTX 3060 Ti | 8 GB | Fits | F16 | 2.5 GB | 234 | details |
| RTX 3070 | 8 GB | Fits | F16 | 2.5 GB | 234 | details |
| RTX 3080 10GB | 10 GB | Fits | F16 | 2.5 GB | 396 | details |
| RTX 3080 12GB | 12 GB | Fits | F16 | 2.5 GB | 476 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 2.5 GB | 526 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | F16 | 2.5 GB | 150 | details |
| RTX 4070 Super | 12 GB | Fits | F16 | 2.5 GB | 263 | details |
| RTX 4070 Ti | 12 GB | Fits | F16 | 2.5 GB | 263 | details |
| RTX 4070 Ti Super | 16 GB | Fits | F16 | 2.5 GB | 351 | details |
| RTX 4080 | 16 GB | Fits | F16 | 2.5 GB | 374 | details |
| RTX 4080 Super | 16 GB | Fits | F16 | 2.5 GB | 384 | details |
| RTX 5060 | 8 GB | Fits | F16 | 2.5 GB | 234 | details |
| RTX 5070 | 12 GB | Fits | F16 | 2.5 GB | 351 | details |
| RTX 5080 | 16 GB | Fits | F16 | 2.5 GB | 501 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 2.5 GB | 488 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 2.5 GB | 526 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 2.5 GB | 935 | details |
| RTX A6000 | 48 GB | Fits | F16 | 2.5 GB | 401 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 2.5 GB | 501 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 2.5 GB | 935 | details |
| L40S | 48 GB | Fits | F16 | 2.5 GB | 451 | details |
| A100 40GB | 40 GB | Fits | F16 | 2.5 GB | 811 | details |
| A100 80GB | 80 GB | Fits | F16 | 2.5 GB | 1064 | details |
| H100 PCIe | 80 GB | Fits | F16 | 2.5 GB | 1043 | details |
| H100 SXM | 80 GB | Fits | F16 | 2.5 GB | 1747 | details |
| H200 | 141 GB | Fits | F16 | 2.5 GB | 2504 | details |
| RX 7600 | 8 GB | Fits | F16 | 2.5 GB | 150 | details |
| RX 7700 XT | 12 GB | Fits | F16 | 2.5 GB | 225 | details |
| RX 7800 XT | 16 GB | Fits | F16 | 2.5 GB | 325 | details |
| RX 7900 GRE | 16 GB | Fits | F16 | 2.5 GB | 300 | details |
| RX 7900 XT | 20 GB | Fits | F16 | 2.5 GB | 417 | details |
| RX 9070 | 16 GB | Fits | F16 | 2.5 GB | 334 | details |
| RX 9070 XT | 16 GB | Fits | F16 | 2.5 GB | 334 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 2.5 GB | 134 | details |
| Arc A770 16GB | 16 GB | Fits | F16 | 2.5 GB | 292 | details |
| Mac M1 16GB | 16 GB | Fits | F16 | 2.5 GB | 35 | details |
| Mac M2 24GB | 24 GB | Fits | F16 | 2.5 GB | 52 | details |
| Mac M3 24GB | 24 GB | Fits | F16 | 2.5 GB | 52 | details |
| Mac M4 16GB | 16 GB | Fits | F16 | 2.5 GB | 63 | details |
| Mac M4 32GB | 32 GB | Fits | F16 | 2.5 GB | 63 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | F16 | 2.5 GB | 104 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 2.5 GB | 78 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | F16 | 2.5 GB | 142 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 2.5 GB | 142 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 2.5 GB | 209 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 2.5 GB | 209 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 2.5 GB | 209 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 2.5 GB | 285 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 2.5 GB | 417 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 2.5 GB | 427 | details |
Questions
How much VRAM does PaddleOCR VL 1.6 need?
About 1.3 GB at 4-bit (Q4_K_M) with an 8K context: 0.5 GB of weights (estimated), 0.1 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 1.7 GB.
What is the smallest GPU that runs PaddleOCR VL 1.6?
At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).
Can I run PaddleOCR VL 1.6 on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Qwen3.5 0.8B
873M · 1.5 GB at 4-bitLightOnOCR 3 0.8B
853M · 1.5 GB at 4-bitHunyuanOCR
1.1B · 2.0 GB at 4-bitFastVLM 0.5B
759M · 1.2 GB at 4-bitQwen3 0.6B
752M · 1.9 GB at 4-bitLFM2.5 1.2B Instruct
1.2B · 1.3 GB at 4-bitModel facts
| Hugging Face repo | PaddlePaddle/PaddleOCR-VL-1.6 |
| Parameters | 958,588,736 (959M) |
| Layers / hidden size | 18 / 1024 |
| Attention | GQA; 16 heads, 2 KV heads, head dim 128 |
| KV cache per token (FP16) | 18.0 KiB |
| Max context | 131,072 tokens |
| License | apache-2.0 |
| Released on HF | May 27, 2026 |
| Downloads (30d) / likes | 98,553 / 593 |
Sources: Hugging Face model API · config.json (PaddlePaddle/PaddleOCR-VL-1.6) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.