HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks

granite-vision-3.2-2b

unsloth/granite-vision-3.2-2b

kind
original
quant
bf16 (16-bit)
family
misc
modality
image-text-to-text
root model
ibm-granite/granite-vision-3.2-2b
provider
unsloth
license
apache-2.0
language
en
context
128k
library
transformers
vRAM (GB)
—
size on disk
—
downloads
79
likes
5
created
—
last modified
2025-03-14T09:45:16+00:00 (2025)
huggingface
https://huggingface.co/unsloth/granite-vision-3.2-2b
tags
transformerssafetensorsllava_nextimage-text-to-textconversationalenarxiv:2502.09927base_model:ibm-granite/granite-vision-3.2-2bbase_model:finetune:ibm-granite/granite-vision-3.2-2blicense:apache-2.0text-generation-inferenceendpoints_compatibleregion:us

Benchmark scores (3)

benchmarkcategoryscorecontextconfevidence
DocVQAmultimodal66.0DocVQAmedium| DocVQA | 0.66 | 0.87 | 0.87 | 0.88 | **0.89** |
ChartQAmultimodal60.0ChartQAmedium| ChartQA | 0.60 | 0.75 | 0.81 | 0.82 | **0.87** |
MMMUmultimodal32.0MMMUmedium| MMMU | 0.32 | 0.35 | 0.42 | **0.44** | 0.37 |

Discussions (0)

none scraped.

Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.