Skip to content

MonkeyOCRv2 B Parsing VRAM requirements

How much VRAM MonkeyOCRv2 B Parsing needs
4-bit weights: 508 MBWith 8K context: 2.1 GBWith 32K context: 4.8 GB8-bit, 8K context: 2.5 GB
Parameters
878M
Longest context
40K
Released
Jul 2026
License
Apache 2.0
Strongest at
Text and images

How much VRAM MonkeyOCRv2 B Parsing needs

Memory MonkeyOCRv2 B Parsing needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 40K
16-bit1.6 GB3.3 GB5.9 GB6.8 GB
8-bit890 MB2.5 GB5.1 GB6.0 GB
6-bit687 MB2.3 GB4.9 GB5.8 GB
5-bit596 MB2.2 GB4.9 GB5.7 GB
4-bit508 MB2.1 GB4.8 GB5.6 GB
3-bit409 MB2.0 GB4.7 GB5.5 GB
2-bit314 MB2.0 GB4.6 GB5.5 GB
1.6-bit173 MB1.8 GB4.4 GB5.3 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run MonkeyOCRv2 B Parsing

The best way MonkeyOCRv2 B Parsing runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit40K174 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4060 report
RTX 50608-bit40K262 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5060 report
RTX 30708-bit40K262 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 3070 report
RTX 3080 10 GB8-bit40K386 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 3080 10 GB report
RTX 3060 12 GB8-bit40K220 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 3060 12 GB report
Arc B5808-bit40K173 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Arc B580 report
RTX 40708-bit40K287 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4070 report
RTX 4070 Super8-bit40K287 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4070 Super report
RTX 50708-bit40K354 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5070 report
RTX 4060 Ti 16 GB8-bit40K183 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit40K262 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit40K200 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Arc A770 16 GB report
Radeon RX 9070 XT8-bit40K261 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit40K354 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4070 Ti Super report
RTX 4080 Super8-bit40K377 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4080 Super report
RTX 5070 Ti8-bit40K430 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5070 Ti report
RTX 50808-bit40K449 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5080 report
Radeon RX 7900 XTX8-bit40K331 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Radeon RX 7900 XTX report
RTX 30908-bit40K442 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 3090 report
RTX 40908-bit40K463 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 4090 report
Radeon AI PRO R97008-bit40K261 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Radeon AI PRO R9700 report
RTX 50908-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 5090 report
2× RTX 30908-bit40K442 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 2× RTX 3090 report
2× RTX 40908-bit40K463 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit40K546 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit40K449 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX 6000 Ada report
L40S8-bit40K420 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on L40S report
2× RTX 50908-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 2× RTX 5090 report
4× RTX 30908-bit40K442 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit40K85 tok/s1.1 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit40K85 tok/s1.1 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit40K85 tok/s1.1 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit40K156 tok/s0.5 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit40K156 tok/s0.5 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit40K156 tok/s0.5 sPDF: MonkeyOCRv2 B Parsing on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit40K221 tok/s0.3 sPDF: MonkeyOCRv2 B Parsing on M4 Max 128 GB report
M5 Max 128 GB8-bit40K234 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on M5 Max 128 GB report
M3 Ultra 96 GB8-bit40K250 tok/s0.2 sPDF: MonkeyOCRv2 B Parsing on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit40K250 tok/s0.2 sPDF: MonkeyOCRv2 B Parsing on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit40K250 tok/s0.2 sPDF: MonkeyOCRv2 B Parsing on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit40K291 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit40K134 tok/s0.2 sPDF: MonkeyOCRv2 B Parsing on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit40K134 tok/s0.2 sPDF: MonkeyOCRv2 B Parsing on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit40K189 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on RTX Spark report
DGX Spark8-bit40K175 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on DGX Spark report
Dual DGX Spark8-bit40K175 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on Dual DGX Spark report
DGX Station GB3008-bit40K973 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit40K631 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit40K670 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on A100 80 GB SXM report
H100 SXM8-bit40K808 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on H100 SXM report
4× H200 NVL8-bit40K895 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 4× H200 NVL report
8× H200 NVL8-bit40K895 tok/s0.1 sPDF: MonkeyOCRv2 B Parsing on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf aenawi/MonkeyOCRv2-B-GGUF
vLLM
vllm serve zenosai/MonkeyOCRv2-B-Parsing

Questions

How much VRAM does MonkeyOCRv2 B Parsing need?

MonkeyOCRv2 B Parsing needs 508 MB for its weights at 4-bit, 2.1 GB with an 8K context and 4.8 GB with 32K. At 8-bit, 2.5 GB.

What is the minimum VRAM for MonkeyOCRv2 B Parsing?

About 2.0 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060.

How big is MonkeyOCRv2 B Parsing?

878M parameters: 1.6 GB of weights at 16-bit, 508 MB at 4-bit.

Can MonkeyOCRv2 B Parsing run on an RTX 4090?

Yes. At 8-bit it holds 40K of context and writes about 463 tokens a second.

More models