RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run DeepSeek OCR 2 locally?

DeepSeek OCR 2 by deepseek-ai has 3.4B parameters (1.2B active per token, 64 experts, 6 used per token). At 4-bit with an 8K context it needs about 2.9 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.

2.9 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
8Kmax context
MHAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from sabafallah/DeepSeek-OCR-2-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx
2-bit (Q2_K)rough2.1 GB2.3 GB
3-bit (Q3_K_M)usable2.4 GB2.7 GB
4-bit (Q4_K_M) •good, most popular2.7 GB2.9 GB
5-bit (Q5_K_M)very good3.1 GB3.4 GB
6-bit (Q6_K)near lossless3.5 GB3.7 GB
8-bit (Q8_0) •effectively lossless3.8 GB4.0 GB
16-bit (F16/BF16) •original weights6.4 GB6.6 GB

Which GPUs and Macs run DeepSeek OCR 2?

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).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsF166.6 GB174details
RTX 4060 Ti 16GB16 GBFitsF166.6 GB139details
RTX 309024 GBFitsF166.6 GB452details
RTX 409024 GBFitsF166.6 GB487details
RTX 509032 GBFitsF166.6 GB866details
RTX 5070 Ti16 GBFitsF166.6 GB433details
RX 7900 XTX24 GBFitsF166.6 GB464details
Mac M4 Pro 48GB48 GBFitsF166.6 GB132details
Mac M4 Max 128GB128 GBFitsF166.6 GB264details
Mac M3 Ultra 512GB512 GBFitsF166.6 GB396details
DGX Spark128 GBFitsF166.6 GB132details
RTX 407012 GBFitsF166.6 GB243details
RTX 5060 Ti 16GB16 GBFitsF166.6 GB216details
RTX 40608 GBFitsF166.6 GB131details
Arc B58012 GBFitsF166.6 GB220details
RTX 3060 Ti8 GBFitsF166.6 GB216details
RTX 30708 GBFitsF166.6 GB216details
RTX 3080 10GB10 GBFitsF166.6 GB367details
RTX 3080 12GB12 GBFitsF166.6 GB441details
RTX 3090 Ti24 GBFitsF166.6 GB487details
RTX 4060 Ti 8GB8 GBFitsF166.6 GB139details
RTX 4070 Super12 GBFitsF166.6 GB243details
RTX 4070 Ti12 GBFitsF166.6 GB243details
RTX 4070 Ti Super16 GBFitsF166.6 GB325details
RTX 408016 GBFitsF166.6 GB346details
RTX 4080 Super16 GBFitsF166.6 GB355details
RTX 50608 GBFitsF166.6 GB216details
RTX 507012 GBFitsF166.6 GB325details
RTX 508016 GBFitsF166.6 GB464details
2x RTX 30902×24 GBFitsF166.6 GB452details
2x RTX 40902×24 GBFitsF166.6 GB487details
2x RTX 50902×32 GBFitsF166.6 GB866details
RTX A600048 GBFitsF166.6 GB371details
RTX 6000 Ada48 GBFitsF166.6 GB464details
RTX PRO 6000 Blackwell96 GBFitsF166.6 GB866details
L40S48 GBFitsF166.6 GB417details
A100 40GB40 GBFitsF166.6 GB751details
A100 80GB80 GBFitsF166.6 GB985details
H100 PCIe80 GBFitsF166.6 GB966details
H100 SXM80 GBFitsF166.6 GB1618details
H200141 GBFitsF166.6 GB2318details
RX 76008 GBFitsF166.6 GB139details
RX 7700 XT12 GBFitsF166.6 GB209details
RX 7800 XT16 GBFitsF166.6 GB301details
RX 7900 GRE16 GBFitsF166.6 GB278details
RX 7900 XT20 GBFitsF166.6 GB386details
RX 907016 GBFitsF166.6 GB309details
RX 9070 XT16 GBFitsF166.6 GB309details
Ryzen AI Max+ 395 128GB128 GBFitsF166.6 GB124details
Arc A770 16GB16 GBFitsF166.6 GB270details
Mac M1 16GB16 GBFitsF166.6 GB33details
Mac M2 24GB24 GBFitsF166.6 GB48details
Mac M3 24GB24 GBFitsF166.6 GB48details
Mac M4 16GB16 GBFitsF166.6 GB58details
Mac M4 32GB32 GBFitsF166.6 GB58details
Mac M1 Pro 32GB32 GBFitsF166.6 GB97details
Mac M3 Pro 36GB36 GBFitsF166.6 GB72details
Mac M4 Pro 24GB24 GBFitsF166.6 GB132details
Mac M4 Pro 64GB64 GBFitsF166.6 GB132details
Mac M1 Max 64GB64 GBFitsF166.6 GB193details
Mac M2 Max 96GB96 GBFitsF166.6 GB193details
Mac M3 Max 128GB128 GBFitsF166.6 GB193details
Mac M4 Max 64GB64 GBFitsF166.6 GB264details
Mac M2 Ultra 192GB192 GBFitsF166.6 GB386details
Mac M3 Ultra 256GB256 GBFitsF166.6 GB396details

Questions

How much VRAM does DeepSeek OCR 2 need?

About 2.9 GB at 4-bit (Q4_K_M) with an 8K context: 1.8 GB of weights (real GGUF file size), 0.5 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 4.0 GB.

What is the smallest GPU that runs DeepSeek OCR 2?

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 DeepSeek OCR 2 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 repodeepseek-ai/DeepSeek-OCR-2
Parameters3,389,119,360 (3.4B) · ~1.2B active (computed from config)
Layers / hidden size12 / 1280
AttentionMHA; 10 heads, 10 KV heads, head dim 128
KV cache per token (FP16)60.0 KiB
Max context8,192 tokens
Licenseapache-2.0
Released on HFJan 27, 2026
Downloads (30d) / likes265,018 / 1,156

Sources: Hugging Face model API · config.json (deepseek-ai/DeepSeek-OCR-2) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.