What GPU do I need to run deepseek-ai/DeepSeek-OCR?
A 3.3B (MoE) vision-language model that reads images alongside text. 3.3B parameters, published in BF16. View on Hugging Face
DeepSeek-OCR is published by deepseek-ai on Hugging Face, with 2,472,241 downloads and 3,350 likes to date. It's a DeepseekOCRForCausalLM model built for image-text-to-text, published natively in BF16.
What DeepSeek-OCR is
DeepSeek-OCR is a 3.3B-parameter mixture-of-experts vision-language model published by DeepSeek on Hugging Face. It is released under MIT.
License note: permissive: allows commercial use, modification and redistribution. Facts in this section are sourced from DeepSeek-OCR's Hugging Face model card, not benchmarked by Aquanode.
What it's used for
- Image understanding and captioning
- Visual question answering
- Document/OCR-style reading
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 6.2 GB | 7.5 GB | RTX 5060 Ti | 1 | $0.110/hr |
| FP8 (quantized) | 3.1 GB | 3.7 GB | RTX 5060 Ti | 1 | $0.110/hr |
| INT4 (quantized) | 1.6 GB | 1.9 GB | RTX 5060 Ti | 1 | $0.110/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run DeepSeek-OCR at its published (BF16) precision: 1× RTX 5060 Ti, at $0.110/hr per GPU ($0.110/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
DeepSeek-OCR: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Multi-head attention: every layer caches keys and values for every head. Cached per token: 61,440 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.23 GB |
| 8K (model maximum) | 0.47 GB |
Computed from the layer, head and window counts in the model's own configuration (deepseek-ai/DeepSeek-OCR), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
DeepSeek-OCR: common questions
Does DeepSeek-OCR fit on a 8 GB GPU?
Yes. At BF16 it needs 7.5 GB of VRAM, so an 8 GB card holds it with 0.5 GB to spare. A 6 GB card is not enough for it at BF16.
How many copies of DeepSeek-OCR fit on one RTX 5060 Ti?
2, by VRAM alone. That card carries 16.0 GB and one copy needs 7.5 GB at BF16, on a live rate of $0.110/hr for the whole card. Throughput is not modelled here, so 2 copies is not 2 times the requests served.
What is the least VRAM DeepSeek-OCR can run in?
1.9 GB, at INT4 (quantized), which fits a 6 GB card, against 7.5 GB at BF16. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.
How to run DeepSeek-OCR
Run DeepSeek-OCR with vLLM
Generic example, not from the model's own docs: adjust flags (quantization, context length, parallelism) for your setup.
vllm serve deepseek-ai/DeepSeek-OCR --tensor-parallel-size 1Run DeepSeek-OCR with Ollama
Verified against Ollama's own library listing.
ollama run deepseek-ocrDeploy DeepSeek-OCR on Aquanode
Aquanode has no one-click deploy template for DeepSeek-OCR; you install the inference engine yourself with the commands below. Aquanode sells GPU pods billed per second, not a hosted inference API.
- Launch a bare GPU pod sized to the requirement above (1× RTX 5060 Ti or larger).
- Open a terminal on the pod, or save one of the commands above as a startup script so it runs automatically the first time the pod boots.
- Run the command and connect to the resulting endpoint.
Launch a GPU podWeight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More DeepSeek OCR models
- DeepSeek-OCR-2 (3.4B, BF16)
Alternatives at this size
Other models for image-text-to-text within about a third of DeepSeek-OCR's 3.3B parameters, from other model lines.
- Qwen3.5-4B (4.7B, BF16)
- Unlimited-OCR (3.3B, BF16)
- gemma-3-4b-it (4.3B, BF16)
- Cosmos-Reason2-2B (2.4B, BF16)
- Rax-4.5 (2.3B, BF16)
More on DeepSeek-OCR
Fits on an 8 GB GPU at BF16: every model that fits in 8 GB.
Best vision-language models: how DeepSeek-OCR ranks against the rest.
Best OCR and document parsing models: how DeepSeek-OCR ranks against the rest.
Related reading: H100 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
