Kimi-VL-A3B-Instruct

Home/Models/moonshotai/Kimi-VL-A3B-Instruct

moonshotai/Kimi-VL-A3B-Instruct

kind: original
quant: bf16 (16-bit)
family: kimi
modality: image-text-to-text
root model: moonshotai/Moonlight-16B-A3B
provider: moonshotai
license: mit
language: —
context: —
library: transformers
vRAM (GB): —
size on disk: —
downloads: 379,082
likes: 276
created: —
last modified: 2026-01-30T04:51:04+00:00 (2026)
huggingface: https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct
tags: transformerssafetensorskimi_vlfeature-extractionagentvideoscreenspotlong-contextimage-text-to-textconversationalcustom_codearxiv:2504.07491base_model:moonshotai/Moonlight-16B-A3Bbase_model:finetune:moonshotai/Moonlight-16B-A3Blicense:mitregion:us

benchmark	category	score	context	conf	evidence
MMBench	multimodal	83.1	MMBench-EN-v1.1 (Acc)	medium	`\| MMBench-EN-v1.1 (Acc) \| 83.1 \| 77.1 \| 82.6 \| 65.8 \| 74.6 \| 79.6 \| 83.1 \|`
MathVista	multimodal	63.8	MathVista (Pass@1)	medium	`\| MathVista (Pass@1) \| 63.8 \| 52.5 \| 68.2 \| 47.7 \| 56.1 \| 62.8 \| 68.7 \|`
OSWorld	agentic	5.03	OSWorld (Pass@1)	medium	`\| OSWorld (Pass@1) \| 5.03 \| - \| 2.5 \| - \| - \| - \| 8.22 \|`

none scraped.