MiniCPM-V 4.7 35B-A3B
- Context
- 262Ktokens
- Released
- 1mo ago
- Size
- 0.8Bparameters
LightOnOCR-3-0.8B is the lightest model in LightOn's LightOnOCR-3 family of OCR models, built on the Qwen3.5 vision-language architecture. The maker says the series runs faster and transcribes more accurately than the previous LightOnOCR-2 generation, and adds visual understanding outputs such as labeled bounding boxes, image descriptions, and extracted chart data. The model is built for document transcription and understanding, offered as a single-model alternative to multi-component document pipelines.
- Multimodal model
- Open weights · 0.8B
- Accepts images
How it compares
Additional source evidence
- Source model
- lightonai/lightonocr-3-0.8b
- Last updated
- 2026-10-14
- Model context
- 262K
- Model output limit
- 16K
- Open weights
- Yes
- File attachments
- Yes
- Reasoning
- No
- Tool calling
- No
- Temperature control
- Yes
Downloadable weights
Published benchmarks
Complete source record (JSON)
{
"id": "lightonai/lightonocr-3-0.8b",
"name": "LightOnOCR-3 0.8B",
"description": "LightOnOCR-3-0.8B is the lightest model of LightOn's LightOnOCR-3 family, with the same layout and visual understanding features as its larger siblings.",
"last_updated": "2026-10-14",
"attachment": true,
"reasoning": false,
"tool_call": false,
"temperature": true,
"modalities": {
"input": [
"text",
"image"
],
"output": [
"text"
]
},
"limit": {
"context": 262144,
"output": 16384
},
"benchmarks": [
{
"name": "olmOCR-bench",
"score": 85.5,
"source": "https://huggingface.co/lightonai/LightOnOCR-3-0.8B",
"metric": "score"
},
{
"name": "ParseBench",
"score": 74.6,
"source": "https://huggingface.co/lightonai/LightOnOCR-3-0.8B",
"metric": "mean"
}
],
"release_date": "2026-10-08",
"open_weights": true,
"license": "apache-2.0",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/lightonai/LightOnOCR-3-0.8B"
}
]
}Frequently asked questions
Running locally
LightOnOCR-3 0.8B needs an estimated 1.1 GB of VRAM at Q4, 1.5 GB at Q8 and 2.3 GB at FP16. These figures are estimates from the 0.8B parameter count, covering weights plus about 15% and 0.5 GB for a short context.
Yes, LightOnOCR-3 0.8B fits on a 24 GB, 16 GB, 12 GB or 8 GB GPU, needing an estimated 2.3 GB of VRAM at FP16.
Yes, the estimated 1.1 GB at Q4 fits the smallest common Mac memory size of 8 GB, assuming macOS lets the GPU use about 75% of unified memory.
Privacy and data
Yes, LightOnOCR-3 0.8B has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window of LightOnOCR-3 0.8B is 262K tokens (262,144), and its maximum output is 16K tokens (16,384).
Yes, LightOnOCR-3 0.8B reads text and images. Audio, video and PDFs are not listed as input modalities.
No, LightOnOCR-3 0.8B does not support tool calling (function calling).
No, LightOnOCR-3 0.8B does not support a reasoning or thinking mode.
Quality and benchmarks
The closest alternatives, matched on model kind and release date with no comparable price, are MiniCPM-V 4.7 35B-A3B and MiniCPM-V 4.7 1B by OpenBMB, and Flint by Recursive.
MiniCPM-V 4.7 35B-A3BLightOnOCR-3 0.8B vs MiniCPM-V 4.7 35B-A3BMiniCPM-V 4.7 1BLightOnOCR-3 0.8B vs MiniCPM-V 4.7 1BFlintLightOnOCR-3 0.8B vs Flint
Licensing and access
LightOnOCR-3 0.8B has open weights and is released under the Apache 2.0 license.
Yes, LightOnOCR-3 0.8B can be used commercially because the Apache 2.0 license is permissive and allows commercial use.
The weights of LightOnOCR-3 0.8B are available in the Hugging Face repository lightonai/LightOnOCR-3-0.8B.
Basics
LightOnOCR-3 0.8B was made by Lightonai and released on October 8, 2026.
Yes, LightOnOCR-3 0.8B is free to use, as it is released with open weights.
Generated from Vertical Tabs data, updated November 11, 2026.
Similar models
MiniCPM-V 4.7 35B-A3BMiniCPM-V 4.7 1BMiniCPM-V 4.7 1B
40BFlint
Qwen3.8 Flash Next
36BApodex 1.1 Mini
Ornith 1.5 35B A3BOrnith 1.5 35B A3B
78.1BKolibri 1
120BNemotron 3.5 Super 120B A12B
36BMacaron V1 Tall
Dots3-Note PreviewDots3-Note Preview
ERNIE-5.5-Flash-PreviewERNIE-5.5-Flash-Preview
Ornith 1.0 31B
Discussion
No comments yet
Be the first to share your thoughts.