Please refer to the official vLLM recipe for deployment details:
MiniCPM-OCR-Jev
openbmb·Instruct·released 23 Oct 2026
ollama run hf.co/openbmb/MiniCPM-OCR-Jev-gguf:Q4_K_MMiniCPM-OCR Jev (2B) is an OCR quality-assessment model for text, formula, and table regions. Given a region image and a candidate OCR result, it returns a score from 0 to 1 — higher means the candidate is more likely correct.
How much VRAM does MiniCPM-OCR-Jev need?
At FP16 / BF16 MiniCPM-OCR-Jev needs about 5.5 GB of GPU memory.
| Quantisation | Total |
|---|---|
| 4-bitRecommended default for local use | ≈1.9 GB |
| 8-bitLossless in practice, twice the size of Q4 | ≈2.7 GB |
| FP16 / BF16recommendedFull precision, as released | ≈4.7 GB |
Estimates for the weights plus runtime overhead; activations vary with resolution and batch size. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.
Run MiniCPM-OCR-Jev locally
--quantQ4_K_Mdefault; pick your GPU to tailor itopenbmb/MiniCPM-OCR-Jev-gguf
ollama run hf.co/openbmb/MiniCPM-OCR-Jev-gguf:Q4_K_M llama-server -hf openbmb/MiniCPM-OCR-Jev-gguf:Q4_K_MDownloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.
lms get openbmb/MiniCPM-OCR-Jev-gguf@Q4_K_MThen load it from the LM Studio app or with `lms load`.
vllm serve openbmb/MiniCPM-OCR-Jev vllm serve openbmb/MiniCPM-OCR-Jev-AWQservices:
vllm:
image: vllm/vllm-openai:latest
command: openbmb/MiniCPM-OCR-Jev
ports:
- "8000:8000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 8000.
sglang serve openbmb/MiniCPM-OCR-Jev --host 0.0.0.0 --port 30000services:
sglang:
image: lmsysorg/sglang:latest
command: sglang serve openbmb/MiniCPM-OCR-Jev --host 0.0.0.0 --port 30000
ports:
- "30000:30000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 30000.
Quantised versions and downloads
| Repository | Format | Quant | Size |
|---|---|---|---|
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | IQ3_XS | 1.0 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | IQ3_M | 1.1 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | IQ4_XS | 1.2 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q4_0 | 1.2 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q4_K_S | 1.2 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | IQ4_NL | 1.2 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q4_K_M | 1.3 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q4_1 | 1.3 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q5_0 | 1.4 GB |
| openbmb/MiniCPM-OCR-Jev-gguf | GGUF | Q5_K_S | 1.4 GB |
About MiniCPM-OCR-Jev
MiniCPM-OCR Jev (2B) is an OCR quality-assessment model for text, formula, and table regions. Given a region image and a candidate OCR result, it returns a score from 0 to 1 — higher means the candidate is more likely correct. MiniCPM-OCR Jev is a 2B OCR quality-assessment model for text, formula, and table regions. Given a region image and a candidate OCR result, it returns a score from 0 to 1 — higher means the candidate is more likely correct. Use it to auto-accept reliable regions, route the rest to human review, or pick the best result among several OCR models. • 🛡️ A small model that beats larger ones: at only 2B parameters, decision AUROC (×100) on • JevOCRBench is 93.6, above • JEV-35B-A3B (90.1) and close to Qwen3.5-397B-A17B (93.8). • 🧩 Stable generalization: on parses from MiniCPM-OCR 1.0, GLM-OCR, PaddleOCR-VL-1.6, and • Xiaomi-OCR-0, AUROC (×100) stays in 91.9–93.4. Trained from Qwen3.5-2B. The vision encoder and language backbone are retained, and the language-model output head lm_head is replaced with a binary classification head that learns whether a candidate is correct. When calling the cloud API, the Docker image, or the SDK service with a built-in renderer, supply the original region image, the candidate text, and the region category (text / formula / table) — the service renders internally. A service without a renderer requires you to supply the rendered image as well.
Full model card on Hugging Face →Alternatives to MiniCPM-OCR-Jev
All alternatives →Sparse Reference Sliding Window Attention (SR-SWA): Compresses text-visual attention interactions from ultra-long sequences down to a fixed-size set of working references (e.g., 512 tokens).
Refer to 🌟GitHub for guidance on model inference acceleration and PDF processing, etc.
GLM-OCR is a multimodal OCR model for complex document understanding, built on the GLM-V encoder–decoder architecture. It introduces Multi-Token Prediction (MTP) loss and stable full-task reinforcement learning to improve training efficiency, recognition accuracy, and generalization.
MiniCPM-OCR 1.0 (1.3B) is the first OCR-specific model in the MiniCPM-V/o family. It focuses on text-centric document understanding — text, tables, formulas, charts, seals, and structured information extraction — and balances quality against speed on real-world captures.
We present dots.mocr. Beyond achieving state-of-the-art (SOTA) performance in standard multilingual document parsing among models of comparable size, dots.mocr excels at converting structured graphics (e.g., charts, UI layouts, scientific figures and etc.) directly into SVG code.
MiniCPM-OCR-Jev FAQ
- How much VRAM does MiniCPM-OCR-Jev need?
- About 5.5 GB at FP16 / BF16.
- Can MiniCPM-OCR-Jev run on the RTX 4060?
- Yes. It fits at FP16 / BF16 (about 4.7 GB). It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
- Can I use MiniCPM-OCR-Jev commercially?
- Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.
- How do I run MiniCPM-OCR-Jev with Ollama or llama.cpp?
- With Ollama: ollama run hf.co/openbmb/MiniCPM-OCR-Jev-gguf:Q4_K_M. With llama.cpp: llama-server -hf openbmb/MiniCPM-OCR-Jev-gguf:Q4_K_M, which starts an OpenAI-compatible server.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments