MI

MiniCPM OCR Jev

MiniCPMGGUF by openbmbUpdated 2026-10-29
Download on Hugging Face
Parameters1.88B
Context256K
Layers24
Architectureqwen35
Smallest quant1.0 GB
Quick answer

MiniCPM OCR Jev needs about 2.2 GB of memory at Q4_K_M (a 1.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 15 quants, from 1.0 GB (IQ3_XS) to 3.8 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.

MiniCPM OCR Jev – 1.88B parameter OCR correctness decision model

MiniCPM OCR Jev is a 1.88B parameter model that evaluates OCR results for text, formula, and table regions. It returns a score from 0 to 1 indicating the likelihood that a candidate OCR result is correct. The model is trained from Qwen3.5-2B and uses a binary classification head.

Good at

  • OCR correctness scoring
  • text, formula, and table region evaluation
  • binary decision output

Limits

  • No vision support in llama.cpp without renderer
  • Requires 8k context for memory estimates
  • Apache-2.0 license

Best for

  • Auto‑accept reliable OCR regions
  • Route uncertain OCR results to human review
  • Select best OCR output among multiple models

Who it’s for. Developers building OCR pipelines who need a lightweight decision model to filter OCR outputs. The model is suitable for integration with llama.cpp or cloud APIs that support image rendering.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ3_XS1.0 GB 1.9 GB
Fits Get file →
IQ3_M1.1 GB 2.0 GB
Fits Get file →
IQ4_XS1.2 GB 2.1 GB
Fits Get file →
Q4_01.2 GB 2.1 GB
Fits Get file →
Q4_K_S1.2 GB 2.1 GB
Fits Get file →
IQ4_NL1.2 GB 2.1 GB
Fits Get file →
Q4_K_M1.3 GB 2.2 GB
Fits Get file →
Q4_11.3 GB 2.2 GB
Fits Get file →
Q5_01.4 GB 2.3 GB
Fits Get file →
Q5_K_S1.4 GB 2.3 GB
Fits Get file →
Q5_K_M1.4 GB 2.3 GB
Fits Get file →
Q5_11.5 GB 2.4 GB
Fits Get file →
Q6_K1.6 GB 2.5 GB
Fits Get file →
Q8_02.0 GB 2.9 GB
Fits Get file →
F163.8 GB 4.7 GB
Fits Get file →

For 8 GB VRAM, use IQ4_K_M (1.27 GB, ~2.2 GB memory at 8k context) or Q5_K_M (1.41 GB, ~2.3 GB). For 4 GB VRAM, IQ3_M (1.06 GB, ~2 GB) or IQ4_XS (1.2 GB, ~2.1 GB) are suitable. Higher precision quants like Q6_K or Q8_0 require more memory.

What GPU do you need for MiniCPM OCR Jev?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is Q8_0 (2.0 GB file, about 2.9 GB in use).
  • 6 GB VRAM: the best quant that fits is F16 (3.8 GB file, about 4.7 GB in use).

Which MiniCPM OCR Jev quant should you choose?

Q4_K_M is the usual starting point for MiniCPM OCR Jev: it balances file size and quality well. Here it is a 1.3 GB file.

The Q8_0 file is 2.0 GB, about 1.6 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is IQ3_XS at 1.0 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters1.88B
Layers24
Trained context262,144 tokens
KV cache (fp16)48 KB per token

Source

Quantized byopenbmb
FamilyMiniCPM
Licenseapache-2.0
Downloads (HF)8,132
Last updated2026-10-29

Frequently asked questions about MiniCPM OCR Jev

How much VRAM does MiniCPM OCR Jev need?

About 2.2 GB at Q4_K_M with an 8K context window. That is the 1.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run MiniCPM OCR Jev on an 8 GB GPU?

Yes, with F16: the 3.8 GB file needs about 4.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of MiniCPM OCR Jev GGUF?

It depends on the quant: IQ3_XS is 1.0 GB, IQ3_M is 1.1 GB, IQ4_XS is 1.2 GB, Q4_0 is 1.2 GB, Q4_K_S is 1.2 GB, IQ4_NL is 1.2 GB and 9 more.

What context length does MiniCPM OCR Jev support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.

Where can I download MiniCPM OCR Jev GGUF?

The files are on Hugging Face in the openbmb/MiniCPM-OCR-Jev-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does MiniCPM OCR Jev use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run MiniCPM OCR Jev on 8 GB VRAM?

Yes, quants like IQ4_K_M or Q5_K_M fit within 8 GB VRAM with estimated memory usage around 2.2–2.3 GB at 8k context.

Does the model support vision input directly in llama.cpp?

No, llama.cpp requires a renderer to provide the rendered image; the model itself does not process raw images.

What license applies to MiniCPM OCR Jev?

The model is released under the Apache License 2.0.

Other GGUFs of MiniCPM-OCR-Jev

Similar models

Browse all →

More from openbmb

All by openbmb →

More MiniCPM models

All MiniCPM →

Other tiny models

All tiny models →