Beating the official open JEV release puts a Korean AI startup at the global frontier of benchmark competition.
TL;DR: VIDRAFT's Darwin ZTC v2 has claimed the overall number-one position on the System One Mosaic Benchmark (S1MB) global leaderboard, outranking 102 competing models. The benchmark result places Darwin ZTC v2 ahead of the official open JEV release, marking a significant milestone for the Korean AI startup. The ranking was confirmed on the S1MB leaderboard hosted on Hugging Face Spaces as of October 9, 2026.
VIDRAFT, a Korean Pre-AGI AI startup, has taken the top spot on the System One Mosaic Benchmark (S1MB) global leaderboard with its Darwin ZTC v2 model, surpassing a field of 102 evaluated models and edging out the official open JEV release. The result, published on October 9, 2026, via the S1MB leaderboard on Hugging Face Spaces, represents the highest overall ranking in the benchmark's current standings and signals a notable leap forward for the company on the international AI evaluation stage.
Darwin ZTC v2, VIDRAFT's latest model submission, secured the first-place position in the overall composite rankings of the S1MB leaderboard. The System One Mosaic Benchmark is a publicly accessible evaluation hosted on Hugging Face Spaces that tracks and compares a broad field of models — 102 in total at the time of this result — across its scoring criteria.
Crucially, Darwin ZTC v2 outperformed the official open JEV model, which serves as a widely recognized reference point within the benchmark community. Surpassing an official open release of this nature carries particular weight, as such models are often treated as strong baselines against which newer entrants are measured. The S1MB leaderboard itself has attracted meaningful community engagement, as reflected by its 59 likes on Hugging Face, suggesting it is an actively followed evaluation resource among AI researchers and practitioners.
No additional numerical scores or sub-category breakdowns beyond the overall first-place ranking are confirmed in the available source data, so this article limits its reporting accordingly.
Leaderboard performance on publicly auditable benchmarks like S1MB matters for several reasons. First, it provides a transparent, community-verified signal of model capability that is harder to dismiss than internally reported metrics. When a relatively young startup like VIDRAFT claims the top position across more than 100 models — including established open reference releases — it draws the attention of researchers, developers, and potential partners who track these rankings closely.
Second, the fact that Darwin ZTC v2 overtook the official open JEV release is symbolically and practically significant. Official open releases from established efforts are commonly used as gold-standard baselines. Clearing that bar in an overall composite ranking suggests that VIDRAFT's engineering and research direction is producing models competitive at a global level, not merely within a regional or niche context.
Third, the Hugging Face Spaces platform is one of the most widely used venues for AI model discovery and evaluation in the global research community. A number-one ranking on a leaderboard hosted there carries meaningful visibility, exposing VIDRAFT's work to an international audience that might otherwise be unfamiliar with Korean AI development efforts. For a Pre-AGI focused company with ambitions beyond the Korean market, that kind of organic global exposure is difficult to replicate through conventional marketing channels.
The result also contributes to a broader narrative: Korean AI companies are increasingly showing up at the top of globally contested benchmarks, and VIDRAFT's Darwin ZTC v2 is now part of that story.
Q: What is the S1MB leaderboard and why is a top ranking significant?
A: The System One Mosaic Benchmark (S1MB) is a publicly accessible model evaluation hosted on Hugging Face Spaces that ranks AI models based on composite performance criteria. Topping its leaderboard across 102 models signals competitive capability recognized by the broader global AI research community.
Q: Which model did VIDRAFT use to claim first place on S1MB?
A: VIDRAFT submitted Darwin ZTC v2, which achieved the overall number-one ranking on the S1MB global leaderboard as of October 9, 2026.
Q: What is the official open JEV release, and why does beating it matter?
A: The official open JEV release is a model that serves as a well-recognized baseline on the S1MB benchmark. Darwin ZTC v2 surpassing it in the overall composite ranking indicates that VIDRAFT's model outperformed a standard reference point used by the benchmark community to measure competitive submissions.
Source: S1MB 리더보드 (글로벌) (2026-10-09) — original article