Can I run Qianfan OCR locally?
Qianfan OCR by baidu has 4.7B parameters. At 4-bit with an 8K context it needs about 4.4 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.
Image input needs the vision projector file too: about 0.6 GB more (Reza2kn/Qianfan-OCR-GGUF), not included in the figures below, which are for text-only use.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from Reza2kn/Qianfan-OCR-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx |
|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 2.9 GB | 3.5 GB | 6.9 GB |
| 3-bit (Q3_K_M) • | usable | 3.3 GB | 3.9 GB | 7.3 GB |
| 4-bit (Q4_K_M) • | good, most popular | 3.8 GB | 4.4 GB | 7.7 GB |
| 5-bit (Q5_K_M) • | very good | 4.2 GB | 4.8 GB | 8.1 GB |
| 6-bit (Q6_K) • | near lossless | 4.6 GB | 5.2 GB | 8.6 GB |
| 8-bit (Q8_0) • | effectively lossless | 5.6 GB | 6.2 GB | 9.6 GB |
| 16-bit (F16/BF16) • | original weights | 9.5 GB | 10.1 GB | 13.4 GB |
Which GPUs and Macs run Qianfan OCR?
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.) | Details |
|---|---|---|---|---|---|---|
| RTX 3060 12GB | 12 GB | Fits | F16 | 10.1 GB | 41 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | F16 | 10.1 GB | 33 | details |
| RTX 3090 | 24 GB | Fits | F16 | 10.1 GB | 106 | details |
| RTX 4090 | 24 GB | Fits | F16 | 10.1 GB | 114 | details |
| RTX 5090 | 32 GB | Fits | F16 | 10.1 GB | 203 | details |
| RTX 5070 Ti | 16 GB | Fits | F16 | 10.1 GB | 101 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 10.1 GB | 109 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 10.1 GB | 31 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 10.1 GB | 62 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 10.1 GB | 93 | details |
| DGX Spark | 128 GB | Fits | F16 | 10.1 GB | 31 | details |
| RTX 4070 | 12 GB | Fits | F16 | 10.1 GB | 57 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | F16 | 10.1 GB | 51 | details |
| RTX 4060 | 8 GB | Fits | Q8_0 | 6.2 GB | 58 | details |
| Arc B580 | 12 GB | Fits | F16 | 10.1 GB | 52 | details |
| RTX 3060 Ti | 8 GB | Fits | Q8_0 | 6.2 GB | 95 | details |
| RTX 3070 | 8 GB | Fits | Q8_0 | 6.2 GB | 95 | details |
| RTX 3080 10GB | 10 GB | Fits | Q8_0 | 6.2 GB | 162 | details |
| RTX 3080 12GB | 12 GB | Fits | F16 | 10.1 GB | 103 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 10.1 GB | 114 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | Q8_0 | 6.2 GB | 61 | details |
| RTX 4070 Super | 12 GB | Fits | F16 | 10.1 GB | 57 | details |
| RTX 4070 Ti | 12 GB | Fits | F16 | 10.1 GB | 57 | details |
| RTX 4070 Ti Super | 16 GB | Fits | F16 | 10.1 GB | 76 | details |
| RTX 4080 | 16 GB | Fits | F16 | 10.1 GB | 81 | details |
| RTX 4080 Super | 16 GB | Fits | F16 | 10.1 GB | 83 | details |
| RTX 5060 | 8 GB | Fits | Q8_0 | 6.2 GB | 95 | details |
| RTX 5070 | 12 GB | Fits | F16 | 10.1 GB | 76 | details |
| RTX 5080 | 16 GB | Fits | F16 | 10.1 GB | 109 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 10.1 GB | 106 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 10.1 GB | 114 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 10.1 GB | 203 | details |
| RTX A6000 | 48 GB | Fits | F16 | 10.1 GB | 87 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 10.1 GB | 109 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 10.1 GB | 203 | details |
| L40S | 48 GB | Fits | F16 | 10.1 GB | 98 | details |
| A100 40GB | 40 GB | Fits | F16 | 10.1 GB | 176 | details |
| A100 80GB | 80 GB | Fits | F16 | 10.1 GB | 230 | details |
| H100 PCIe | 80 GB | Fits | F16 | 10.1 GB | 226 | details |
| H100 SXM | 80 GB | Fits | F16 | 10.1 GB | 379 | details |
| H200 | 141 GB | Fits | F16 | 10.1 GB | 543 | details |
| RX 7600 | 8 GB | Fits | Q8_0 | 6.2 GB | 61 | details |
| RX 7700 XT | 12 GB | Fits | F16 | 10.1 GB | 49 | details |
| RX 7800 XT | 16 GB | Fits | F16 | 10.1 GB | 71 | details |
| RX 7900 GRE | 16 GB | Fits | F16 | 10.1 GB | 65 | details |
| RX 7900 XT | 20 GB | Fits | F16 | 10.1 GB | 90 | details |
| RX 9070 | 16 GB | Fits | F16 | 10.1 GB | 72 | details |
| RX 9070 XT | 16 GB | Fits | F16 | 10.1 GB | 72 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 10.1 GB | 29 | details |
| Arc A770 16GB | 16 GB | Fits | F16 | 10.1 GB | 63 | details |
| Mac M1 16GB | 16 GB | Fits | Q8_0 | 6.2 GB | 14 | details |
| Mac M2 24GB | 24 GB | Fits | F16 | 10.1 GB | 11 | details |
| Mac M3 24GB | 24 GB | Fits | F16 | 10.1 GB | 11 | details |
| Mac M4 16GB | 16 GB | Fits | Q8_0 | 6.2 GB | 26 | details |
| Mac M4 32GB | 32 GB | Fits | F16 | 10.1 GB | 14 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | F16 | 10.1 GB | 23 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 10.1 GB | 17 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | F16 | 10.1 GB | 31 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 10.1 GB | 31 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 10.1 GB | 45 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 10.1 GB | 45 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 10.1 GB | 45 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 10.1 GB | 62 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 10.1 GB | 90 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 10.1 GB | 93 | details |
Questions
How much VRAM does Qianfan OCR need?
About 4.4 GB at 4-bit (Q4_K_M) with an 8K context: 2.5 GB of weights (real GGUF file size), 1.1 GB of KV cache and 0.7 GB of runtime overhead. At 8-bit it needs about 6.2 GB.
What is the smallest GPU that runs Qianfan OCR?
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 Qianfan OCR 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.
Unlimited OCR 2
3.3B · 3.0 GB at 4-bitUnlimited OCR
3.3B · 2.9 GB at 4-bitERNIE 4.5 21B A3B Thinking
22B · 13.6 GB at 4-bitMage VL Cowork
4.7B · 4.3 GB at 4-bitQwen3.5 4B
4.7B · 3.9 GB at 4-bitWeVisOne VL 4B
4.8B · 4.3 GB at 4-bitEmbed this
Free badge and "Will it run on my GPU?" widget for model cards, READMEs and blogs. No cookies, no tracking. All options.
Copy a snippet
Markdown badge
[](https://runsonmy.com/model/qianfan-ocr/) Widget (script)
<div class="runsonmy-widget" data-model="qianfan-ocr"><a href="https://runsonmy.com/model/qianfan-ocr/">Will Qianfan OCR run on my GPU? Check on RunsOnMy</a></div>
<script async src="https://runsonmy.com/embed.js"></script> Widget (iframe)
<iframe src="https://runsonmy.com/embed/m/qianfan-ocr/" width="100%" height="220" style="max-width:440px;border:0" loading="lazy" title="Will Qianfan OCR run on my GPU? (RunsOnMy)"></iframe> Model facts
| Hugging Face repo | baidu/Qianfan-OCR |
| Parameters | 4,741,408,256 (4.7B) |
| Layers / hidden size | 36 / 2560 |
| Attention | GQA; 32 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 144.0 KiB |
| Max context | 32,768 tokens |
| License | apache-2.0 |
| Released on HF | Mar 18, 2026 |
| Downloads (30d) / likes | 16,126 / 1,217 |
Sources: Hugging Face model API · config.json (baidu/Qianfan-OCR) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.