Can I run HunyuanOCR locally?
HunyuanOCR by tencent has 1.1B parameters. At 4-bit with an 8K context it needs about 2.0 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.
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.
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 | 1.4 GB | 1.8 GB | 4.0 GB | 13.0 GB |
| 3-bit (Q3_K_M) | usable | 1.5 GB | 1.9 GB | 4.1 GB | 13.1 GB |
| 4-bit (Q4_K_M) | good, most popular | 1.6 GB | 2.0 GB | 4.3 GB | 13.3 GB |
| 5-bit (Q5_K_M) | very good | 1.7 GB | 2.1 GB | 4.4 GB | 13.4 GB |
| 6-bit (Q6_K) | near lossless | 1.9 GB | 2.2 GB | 4.5 GB | 13.5 GB |
| 8-bit (Q8_0) | effectively lossless | 2.1 GB | 2.5 GB | 4.7 GB | 13.7 GB |
| 16-bit (F16/BF16) | original weights | 3.1 GB | 3.5 GB | 5.7 GB | 14.7 GB |
Which GPUs and Macs run HunyuanOCR?
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 | 3.5 GB | 161 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | F16 | 3.5 GB | 129 | details |
| RTX 3090 | 24 GB | Fits | F16 | 3.5 GB | 418 | details |
| RTX 4090 | 24 GB | Fits | F16 | 3.5 GB | 450 | details |
| RTX 5090 | 32 GB | Fits | F16 | 3.5 GB | 800 | details |
| RTX 5070 Ti | 16 GB | Fits | F16 | 3.5 GB | 400 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 3.5 GB | 429 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 3.5 GB | 122 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 3.5 GB | 244 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 3.5 GB | 366 | details |
| DGX Spark | 128 GB | Fits | F16 | 3.5 GB | 122 | details |
| RTX 4070 | 12 GB | Fits | F16 | 3.5 GB | 225 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | F16 | 3.5 GB | 200 | details |
| RTX 4060 | 8 GB | Fits | F16 | 3.5 GB | 121 | details |
| Arc B580 | 12 GB | Fits | F16 | 3.5 GB | 204 | details |
| RTX 3060 Ti | 8 GB | Fits | F16 | 3.5 GB | 200 | details |
| RTX 3070 | 8 GB | Fits | F16 | 3.5 GB | 200 | details |
| RTX 3080 10GB | 10 GB | Fits | F16 | 3.5 GB | 339 | details |
| RTX 3080 12GB | 12 GB | Fits | F16 | 3.5 GB | 407 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 3.5 GB | 450 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | F16 | 3.5 GB | 129 | details |
| RTX 4070 Super | 12 GB | Fits | F16 | 3.5 GB | 225 | details |
| RTX 4070 Ti | 12 GB | Fits | F16 | 3.5 GB | 225 | details |
| RTX 4070 Ti Super | 16 GB | Fits | F16 | 3.5 GB | 300 | details |
| RTX 4080 | 16 GB | Fits | F16 | 3.5 GB | 320 | details |
| RTX 4080 Super | 16 GB | Fits | F16 | 3.5 GB | 329 | details |
| RTX 5060 | 8 GB | Fits | F16 | 3.5 GB | 200 | details |
| RTX 5070 | 12 GB | Fits | F16 | 3.5 GB | 300 | details |
| RTX 5080 | 16 GB | Fits | F16 | 3.5 GB | 429 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 3.5 GB | 418 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 3.5 GB | 450 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 3.5 GB | 800 | details |
| RTX A6000 | 48 GB | Fits | F16 | 3.5 GB | 343 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 3.5 GB | 429 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 3.5 GB | 800 | details |
| L40S | 48 GB | Fits | F16 | 3.5 GB | 386 | details |
| A100 40GB | 40 GB | Fits | F16 | 3.5 GB | 694 | details |
| A100 80GB | 80 GB | Fits | F16 | 3.5 GB | 910 | details |
| H100 PCIe | 80 GB | Fits | F16 | 3.5 GB | 893 | details |
| H100 SXM | 80 GB | Fits | F16 | 3.5 GB | 1496 | details |
| H200 | 141 GB | Fits | F16 | 3.5 GB | 2143 | details |
| RX 7600 | 8 GB | Fits | F16 | 3.5 GB | 129 | details |
| RX 7700 XT | 12 GB | Fits | F16 | 3.5 GB | 193 | details |
| RX 7800 XT | 16 GB | Fits | F16 | 3.5 GB | 279 | details |
| RX 7900 GRE | 16 GB | Fits | F16 | 3.5 GB | 257 | details |
| RX 7900 XT | 20 GB | Fits | F16 | 3.5 GB | 357 | details |
| RX 9070 | 16 GB | Fits | F16 | 3.5 GB | 286 | details |
| RX 9070 XT | 16 GB | Fits | F16 | 3.5 GB | 286 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 3.5 GB | 114 | details |
| Arc A770 16GB | 16 GB | Fits | F16 | 3.5 GB | 250 | details |
| Mac M1 16GB | 16 GB | Fits | F16 | 3.5 GB | 30 | details |
| Mac M2 24GB | 24 GB | Fits | F16 | 3.5 GB | 45 | details |
| Mac M3 24GB | 24 GB | Fits | F16 | 3.5 GB | 45 | details |
| Mac M4 16GB | 16 GB | Fits | F16 | 3.5 GB | 54 | details |
| Mac M4 32GB | 32 GB | Fits | F16 | 3.5 GB | 54 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | F16 | 3.5 GB | 89 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 3.5 GB | 67 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | F16 | 3.5 GB | 122 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 3.5 GB | 122 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 3.5 GB | 179 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 3.5 GB | 179 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 3.5 GB | 179 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 3.5 GB | 244 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 3.5 GB | 357 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 3.5 GB | 366 | details |
Questions
How much VRAM does HunyuanOCR need?
About 2.0 GB at 4-bit (Q4_K_M) with an 8K context: 0.6 GB of weights (estimated), 0.8 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 2.5 GB.
What is the smallest GPU that runs HunyuanOCR?
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 HunyuanOCR 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.
Model facts
| Hugging Face repo | tencent/HunyuanOCR |
| Parameters | 1,119,925,744 (1.1B) |
| Layers / hidden size | 24 / 1024 |
| Attention | GQA; 16 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 96.0 KiB |
| Max context | 131,072 tokens |
| License | other |
| Released on HF | Nov 18, 2025 |
| Downloads (30d) / likes | 640,777 / 847 |
Sources: Hugging Face model API · config.json (tencent/HunyuanOCR) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.