# Datapoint AI releases SVG-Bench

[View media](<https://creatorstoolbox.com/api/storage/native-videos/0cd83d6d6a71783654c45a219a0f926de024d69bf03632259e21226d5330883e.mp4>)

0:000:00

Datapoint AI released [SVG-Bench](<https://x.com/datapointai/status/2113295713211928577>), an open dataset and leaderboard built from 352,270 human preference votes on text-to-SVG generation.

It compares 14 models across eight categories, including icons, logos, illustrations and diagrams. GPT-6 Astra leads the overall board, followed by Gemini 3.8 Flash and Muse Wonder. You can browse the [leaderboard](<https://trydatapoint.com/benchmark/leaderboard?category_id=btrk_e079ae2e1fe7&track_id=btrk_f5061ea0a442>) or get the [dataset on Hugging Face](<https://huggingface.co/datasets/datapointai/text-to-svg-human-preferences-352k>).

Votes, prompts and metadata are CC-BY-4.0 with attribution to Datapoint AI. The SVGs and renders are distributed for research and evaluation under each model provider’s terms, rather than as a curated training corpus. The terms prohibit trying to re-identify annotators.

Loading replies…

Loading replies

---
Canonical source: https://creatorstoolbox.com/@scout/post/n_1dapxfvfgxoug