Press
Global benchmark records
Coverage of results on public leaderboards and international competitions.
In brief
Reports covering our verified #1 in the Google x Hugging Face inference-acceleration challenge, GPQA Diamond, the K-AI leaderboard and Polaris drug-property prediction. Every figure can be checked on the public leaderboard itself.
Government programmes & public sectorGlobal benchmark recordsModels & architectureInference acceleration & servingOn-device & compressionDrug discovery & bioBusiness & partnershipsAll coverage
Global benchmark records coverage
- 2026-09-09IT ChosunVIDRAFT and Pyo Biotech partner on cell-based validation of AI-discovered drug candidates
- 2026-09-07ZDNet Korea“Korea holds just 0.4% of global open-model downloads” — VIDRAFT analyzes Hugging Face top 300
- 2026-09-06Electronic TimesHugging Face spotlights Korea’s VIDRAFT — proving its technology on the global AI stage (CEO Kim Min-sik interview)
- 2026-09-01AI TimesVIDRAFT unveils live experiment world ‘ai-world’: is AI civilization emergence or recitation?
- 2026-09-01BioTimesVIDRAFT AI drug-discovery challenge ‘Open Discovery Challenge’ tops 10,000 submissions
- 2026-09-01JoongAng IlboVIDRAFT AX-RAY adopted to verify safety of national security AI foundation model
- 2026-08-26AI TimesVIDRAFT extends its causal-leakage findings into AX-RAY, offered as the safety-verification core for a national security foundation-model programme
- 2026-08-23vc.ru (Russia)Russian tech outlet vc.ru — Darwin-36B-Opus: bred rather than trained, matching 397B models at 88.4% on GPQA Diamond
- 2026-08-18IT ChosunVIDRAFT releases the AX-RAY AI safety diagnostics leaderboard and evaluation dataset
- 2026-08-18Dong-A IlboVIDRAFT's AI drug discovery challenge passes 2,000 submissions three days after launch
- 2026-08-17AI TimesAI Times — VIDRAFT detects causal leakage in general-purpose LLMs and publishes AX-RAY, an AI safety diagnostic leaderboard and evaluation dataset, on Hugging Face
- 2026-08-15note (Japan)note (Japan) — an in-depth analysis of VIDRAFT's verified #1 record in the Fast Gemma Challenge
- 2026-08-04TLDR AI (US)TLDR AI (US) — VIDRAFT documented the full configuration behind its verified state-of-the-art Fast Gemma submission on a single NVIDIA A10G, featured in the 1.1M-subscriber developer newsletter
- 2026-08-03Korea Economic TVKorea Economic TV — VIDRAFT posts the top verified score on Google×Hugging Face ‘The Fast Gemma Challenge’ (510.58 TPS at PPL 2.39, ~6x baseline)
- 2026-08-03AI TimesAI Times — VIDRAFT takes global #1 on Google’s Fast Gemma Challenge (verified 510.58 TPS at PPL 2.39, ~6x the baseline)
- 2026-08-03ETNewsETNews — VIDRAFT tops the Google × Hugging Face ‘Fast Gemma Challenge’ on verified records
- 2026-08-03Dong-A IlboDong-A Ilbo — Korean AI startup VIDRAFT takes global #1 on Google’s Fast Gemma Challenge (verified records)
- 2026-08-03The Korea Economic Daily (Hankyung)The Korea Economic Daily — VIDRAFT posts the top verified record on the Google × Hugging Face Fast Gemma Challenge
- 2026-08-02techora.ru (Russia)Russian tech outlet techora.ru features VIDRAFT’s Aether-7B-5Attn in a sovereign-AI deep dive — “the world’s first fully open foundation model built by a private company”
- 2026-07-27Dong-A IlboDong-A Ilbo reports — Ourbox reaches #2 in the 30B+ category of the K-AI Leaderboard with its own Korean LLM 'Ourbox-31B-JGOS'; the on-device AI for its talking 3D avatar is built with VIDRAFT's 'POCKET'
- 2026-07-27ZDNet KoreaZDNet Korea reports — Smart logistics firm Ourbox reaches #2 on the K-AI Leaderboard with its own Korean AI 'Ourbox-31B-JGOS', in collaboration with VIDRAFT's CPU on-device platform 'POCKET'
- 2026-07-24Ground News (Canada)Canadian news bias/fact comparison aggregator Ground News aggregates news of Spot robot's Korean voice control at Seoul museum — recorded as listed across 58 news sources including US outlets (74% centrist media)
- 2026-07-13Dong-A IlboKorean AI startup VIDRAFT's 'Darwin' surpasses 1 million global downloads
- 2026-07-09MSNVIDRAFT unveils LLM inference acceleration engine 'VKAE' leaderboard and integrated container
- 2026-07-08Economist KoreaVIDRAFT demonstrates quantum cryptanalysis with IBM 156-qubit quantum computer... verifies symmetric key structure
- 2026-07-08AI TimesVIDRAFT unveils inference acceleration engine 'VKAE' — provides leaderboard and integrated container
- 2026-07-07BrunchFrom Beijing to Moscow — the world reads 'VIDRAFT'
- 2026-07-07Wedoany (China)Global Wedoany (English), introduces VIDRAFT VKAE engine — throughput up to 23.4x
- 2026-07-06hegai.media (Russia)Russian AI business media hegai.media reports on VKAE — evaluates "can be directly verified before adoption via reproducible container"
- 2026-07-06Digital TimesVIDRAFT unveils LLM inference acceleration engine 'VKAE' leaderboard and integrated container
- 2026-07-06Digital DailyVIDRAFT unveils LLM inference acceleration engine 'VKAE' — supports leaderboard and integrated container
- 2026-07-03BehindGenieGenAI·VIDRAFT jointly develop 'metacognition' leaderboard essential for AGI, unveiled
- 2026-07-03BrunchDarwin platform occupies half of the K-AI leaderboard top 20
- 2026-07-02AI TimesGenieGenAI unveils metacognition benchmark, AI model leaderboard, and per-model metacognition adapters simultaneously
- 2026-07-01NateGenieGenAI·VIDRAFT jointly develop 'metacognition' leaderboard essential for AGI, appears
- 2026-07-01LM Studio (US)Darwin-9B-Opus-Code deployed on LM Studio hub — accessible to desktop LLM users worldwide (global distribution)
- 2026-06-29AINEWSDEN (Global)Global AI outlet AINEWSDEN highlights VIDRAFT's 'Darwin Family' — achieving 88.89% on GPQA Diamond without training
- 2026-06-18BehindVIDRAFT records 90.9% on GPQA Diamond··· ranks 1st on Hugging Face's official leaderboard
- 2026-06-17AI TimesVIDRAFT achieves 90.9% on GPQA Diamond… "proves AI scientific reasoning technology"
- 2026-06-17DatanetVIDRAFT's scientific reasoning model 'Darwin-398B-JGOS' ranks No. 1 on Hugging Face leaderboard
- 2026-06-17IT BizVIDRAFT scores 90.9% on AI scientific reasoning benchmark… No. 1 on official Hugging Face leaderboard
- 2026-06-16v.daum.netVIDRAFT scores 90.9% on GPQA Diamond… No. 1 on AI scientific reasoning leaderboard
- 2026-06-16EdailyVIDRAFT's AI model 'Darwin-398B-JGOS' scores 90.9% on GPQA Diamond
- 2026-06-16DailianVIDRAFT's 'AI scientific reasoning language model' ranks No. 1 with 90.9% accuracy on global benchmark
- 2026-06-16NateVIDRAFT's AI model 'Darwin-398B-JGOS' scores 90.9% on GPQA Diamond
- 2026-06-15Electronic TimesVIDRAFT achieves No. 1 in AI scientific reasoning… 'Darwin-398B-JGOS' tops global leaderboard
- 2026-06-15BrunchWe 'crossbred' AI to become world No. 1
- 2026-06-12BehindVIDRAFT's Korean LLM ranks No. 1 overall on K-AI leaderboard, 'proving competitiveness'
- 2026-06-11Economist KoreaVIDRAFT ranks No. 1 overall on K-AI leaderboard with DARWIN-based Korean LLM
- 2026-06-10Hankyung MagazineVIDRAFT's Korean LLM records No. 1 overall on K-AI leaderboard
- 2026-06-09Electronic TimesVIDRAFT ranks No. 1 overall on K-AI leaderboard… DARWIN-based Korean LLM proves competitiveness
- 2026-06-07BrunchThe Darwin platform technology behind the K-AI leaderboard No. 1
- 2026-06-03OpenCSG (China·distribution)Darwin-218B-Delphi distributed on China's open-source AI hub OpenCSG (global distribution)
- 2026-05-19Electronic TimesKorean startup VIDRAFT ranks 3rd globally in AI inference performance without additional GPU training
- 2026-05-16BrunchKorean startup ranks 3rd in world in AI without GPU training
- 2026-04-30v.daum.netHugging Face-certified leaderboard: 16 Chinese models and 5 VIDRAFT (Korea) models among top 21
- 2026-04-28IT ChosunHugging Face-certified leaderboard reorganized around China... Korea's VIDRAFT advances
- 2026-04-28BigGo Finance (Taiwan)Taiwan's BigGo Finance spotlights VIDRAFT's 5 'Darwin' models on HF GPQA leaderboard — all Korean entries are Darwin
- 2026-04-27BrunchHugging Face's GPQA benchmark top 21 survey results shocking
- 2026-04-15BrunchWorld's first multimodality heterogeneous transfer model unveiled
- 2026-04-12BrunchDo AI models interbreed too? — Reaching world No. 5 without training
- 2026-03-12Economist KoreaVIDRAFT globally unveils 'MARL,' an AI reasoning middleware that reduces LLM hallucination
- 2026-03-12BehindVIDRAFT globally unveils reasoning middleware 'MARL' that reduces hallucination
- 2026-03-11Digital TimesAI VIDRAFT globally unveils reasoning middleware MARL that reduces hallucination
- 2025-06-26BrunchK-AI Community Summit 2025
- 2024-12-04BrunchGiniGEN, Korean AI that caught the world's attention
- 2024-08-02Digital Chosun[Interview] Domestic 'on-device adaptive AI' that caught the world's attention
Note
Each entry links to the original article on the publisher's own site. Copyright of the article text belongs to each publisher.