jina-vlm

Home/Models/jinaai/jina-vlm

jinaai/jina-vlm

kind: original
quant: bf16 (16-bit)
family: jina
modality: image-text-to-text
root model: Qwen/Qwen3-1.7B-Base
provider: jinaai
license: cc-by-nc-4.0
language: en, zh, ar, pt, ru, tr, de, es, fr, it, ja, ko
context: —
library: transformers
vRAM (GB): —
size on disk: —
downloads: 529
likes: 119
created: —
last modified: 2026-04-02T16:23:19+00:00 (2026)
huggingface: https://huggingface.co/jinaai/jina-vlm
tags: transformerssafetensorsjvlmtext-generationmultimodalmultilingualvlmvision-languageqwen3siglip2image-text-to-textconversationalcustom_codeenzharptrutrdeesfritja

benchmark	category	score	context	conf	evidence
DocVQA	multimodal	90.6		low	`\| jina-vlm \| 82.0 \| 81.9 \| 83.2 \| 90.6 \| 71.6 \| 778 \| 67.2 \| 32.3/63.5 \| 72.3 \|`
ChartQA	multimodal	81.9		low	`\| jina-vlm \| 82.0 \| 81.9 \| 83.2 \| 90.6 \| 71.6 \| 778 \| 67.2 \| 32.3/63.5 \| 72.3 \|`
MMBench	multimodal	74.3		medium	`MMBench: 74.3`
HellaSwag	knowledge	59.4	GSM-8K	low	`\| jina-vlm \| 56.1 \| 30.3 \| 71.3 \| 77.3 \| 59.4 \|`
MMLU	knowledge	56.1	GSM-8K	low	`\| jina-vlm \| 56.1 \| 30.3 \| 71.3 \| 77.3 \| 59.4 \|`
MMLU-Pro	knowledge	30.3	GSM-8K	low	`\| jina-vlm \| 56.1 \| 30.3 \| 71.3 \| 77.3 \| 59.4 \|`

none scraped.