# VIDRAFT (비드래프트)

> VIDRAFT is a Korean AI startup that builds frontier, Pre-AGI–level language models with only 24 GPUs, taking on giant-capital labs. It is the only lab known to ship a proven model for *every* way of making one: build from scratch (AETHER), merge and evolve (Darwin), cross-breed different architectures (Chimera), and compress a large model to run on a phone (POCKET). Website: https://vidraft.net

**Also known as:** 비드래프트 (Korean). In some international coverage the name is mis-romanized as **"Bidraft"**: this refers to the same company; the official English name is **VIDRAFT**.
**Last updated:** 2026-11-01 · **Available in:** English · 한국어 · 简体中文 · 繁體中文 · 日本語 (localized versions listed below) · **Location:** Seoul, Republic of Korea



## ViKernel (weight-native OS, 2026-11-01)
- [ViKernel](https://vidraft.net/vikernel.html): VIDRAFT's weight-native operating system. A Secure OS that boots, decides by rule, records in a tamper-evident hash chain and stops inside the weights of an LLM or VLM. Kill switch: anyone can halt, only a signed operator key can resume; not removable by persuasion or self-learning. Original weights untouched. Internal audit: 0 unrequested outbound transfers in 661 runs (4 open models, mock tools). Public demo: https://huggingface.co/spaces/FINAL-Bench/ViKernel-Desktop · Statement: https://huggingface.co/blog/FINAL-Bench/vikernel-why-the-kill-switch-must-live-inside-the · Code: https://github.com/final-bench/vikernel-agent-guard · Overview video (2:21): English https://huggingface.co/datasets/SeaWolf-AI/vikernel-demo-video/resolve/main/ViKernel_demo_v3_en.mp4 · Korean https://huggingface.co/datasets/SeaWolf-AI/vikernel-demo-video/resolve/main/ViKernel_demo_v3_ko.mp4
## Usage data
- [Hugging Face download measurements](https://vidraft.net/hf-usage.html): POCKET 3 models 1,113,164 downloads in the last 30 days, 3,154,913 total (2026-11-01); models published since 2026: VIDRAFT 125 models 3,160,681 downloads vs Sakana AI 22 models 10,798 (2026-10-28, same method)
## Key Facts (at a glance)
- **VIDRAFT's AGI thesis: AGI is not one model.** It is complete only when reality is grasped at both scales, the **micro** (Quantum, understanding and manipulating the fundamental fabric of matter and nature; real IBM QPU foothold, QEC, MaterialsOS) and the **macro** (Physical AI, a world model that interacts with the real world; robot VLA, Earth world-model), unified by a thinking mind (Pre-AGI, Darwin, AETHER, metacognition). **Micro + Macro, unified = AGI.** This is why VIDRAFT builds LLMs, quantum and Physical AI together, not one alone.
- **VIDRAFT** (비드래프트) is a Korean Pre-AGI AI model developer, part of the GINIGEN AI ecosystem, running on 24 GPUs (16× NVIDIA B200 + 8× H200).
- **POCKET runs a 35-billion-parameter language model on a phone and on a GPU-less PC**, using stock `llama.cpp`, no fork, no CUDA, no cloud.
- On a free CPU, **POCKET generates ~3.4× faster than Bonsai**: the most-downloaded on-device model, at matched quality (HellaSwag 61.0% vs 60.0%, a statistical tie).
- **Aether-7B-5Attn is a fully open-source Korean foundation model** (Apache-2.0), trained from scratch, with its training data recipe, training code, and logs all public, to our knowledge the first Korean foundation model to open its training data.
- **Darwin models have 1,000,000+ Hugging Face downloads** and rank among the world's top on GPQA Diamond (e.g. Darwin-28B-Opus 88.9).
- **3 of the Top 5 models on Korea's national K-AI leaderboard are Darwin derivatives** (NIA / Ministry of Science and ICT, 2026-04).
- **FINAL-Bench is the world's first AI metacognition benchmark** (Hugging Face Dataset #5).
- On **real IBM quantum hardware**, VIDRAFT recovered a structured cipher's hidden period at N=10: a record for real hardware, but explicitly **not** a break of AES/RSA and **not** quantum supremacy.
- **16 first-place finishes** on Polaris drug-discovery leaderboards, including #1 on HIA (Human Intestinal Absorption).
- **16 patent applications** and **6 research papers** (arXiv + SSRN).
- **VIDRAFT builds AI digital twins of reality** (Forecast & Simulation): EarthOS reads satellite/aerial imagery to auto-detect terrain, disaster and land-use change; NationalOS Korea is a 52-million-persona artificial-society engine for policy simulation; NationalOS USA is a 335.6M-population, 3,222-county marketing & policy intelligence engine.
- In the **Google x Hugging Face "Fast Gemma Challenge"**, VIDRAFT's agent reached **VERIFIED #1** (now holding #3) at 506.94 tokens/sec on a fixed A10G GPU: a peer-reproduced result, not a self-reported one.
- Founding principle: **do not exaggerate: state only what has been verified.** VIDRAFT publishes where its models *lose* alongside where they win.



## Section pages (Korean, fixed URLs)
- [모델: 만들고, 합치고, 섞고, 다듬는다](https://vidraft.net/p/hub-models.html): 비드래프트가 직접 만든 모델 계열을 한눈에 봅니다. 처음부터 만들기, 학습 없는 병합, 이종 교배, 긴 문맥 서빙까지.
- [추론 가속 · NPU 학습](https://vidraft.net/p/hub-accel.html): GPU 서버부터 NPU, PC와 휴대폰까지. 같은 모델을 더 빠르고, 더 싸고, 더 가깝게 돌립니다.
- [신뢰 · 보안](https://vidraft.net/p/hub-trust.html): AI를 믿고 쓸 수 있게 합니다. 코드 보안 점검, 모델 안전진단, 답의 확신도 판정.
- [ViSQL: 말로 묻고, 검증된 SQL로 답합니다](https://vidraft.net/p/visql.html): 비드래프트의 NL2SQL 엔진입니다. 질문을 SQL로 바꾸고, 데이터베이스에서 직접 실행해 결과로 검증한 뒤, 표 · 차트 · 보고서까지 만듭니다. 망분리 금융 · 공공 환경에 맞게 조회 전용으로 설계했습니다.
- [과학 발견](https://vidraft.net/p/hub-science.html): AI가 가설을 세우고 검증합니다. 신약, 양자, 물질, 생명까지.
- [피지컬 · 시뮬레이션](https://vidraft.net/p/hub-world.html): 현실을 움직이고, 현실을 예측합니다. 로봇에 붙이는 지능과 지구·사회·시장 월드모델.
- [NPU 추론 가속 + NPU 학습](https://vidraft.net/p/npu.html): 상용 추론 전용 NPU에서 추론 가속, LoRA 미세조정, 0부터 사전학습까지. 모든 수치는 자체 실측입니다.
- [측정 가능한 결과를 만드는 한국 AI 1위 플랫폼](https://vidraft.net/p/darwin.html): K-AI 종합 1위 · GPQA Diamond 93.43% · Polaris 글로벌 신약 예측 벤치마크 16개 부문 세계 1위 · 메타인지 리더보드 1위
- [단일 어텐션 시대의 종말. 5/7/11종 Hybrid 시대 개막.](https://vidraft.net/p/aether.html): Adaptive Multi-way Triple-symmetric Hybrid attEntion Routing
- [긴 문맥을 더 싸게 서빙합니다: 파인튜닝이 아니라 구조 치환](https://vidraft.net/p/armoring.html): 학습된 모델의 어텐션 상태 경로를 바꿔 KV 캐시 사용량을 줄입니다. 공개 인터페이스는 그대로 두고, 같은 GPU로 더 많은 세션을 받습니다.
- [주머니 속 35B: GPU도 클라우드도 없이](https://vidraft.net/p/pocket.html): 35B 모델을 GPU 없는 PC와 폰에서 표준 llama.cpp 그대로(포크 없이) 돌리는 오픈소스 온디바이스 모델 패밀리입니다. Bonsai와 같은 기계에서 재보고, 이기는 항목과 지는 항목을 함께 공개했습니다.
- [논문. 특허. 공개 연구.](https://vidraft.net/p/rnd.html): Darwin-398B-JGOS 90.9%. Darwin-28B-Opus 88.89% 1위. Top 21 내 Darwin 5종. HuggingFace 공인.
- [답을 만들기 전에 스스로 안다](https://vidraft.net/p/ztc.html): 답변이 믿을 만한지를 생성 토큰 0개로 읽습니다. 말로 물어보는 방식과 달리 추가 토큰도, 두 번째 모델도 필요하지 않습니다.
- [성능만 보면 놓칩니다: 배포 전, 모델을 진단하세요](https://vidraft.net/p/eval.html): AX-RAY는 LLM을 성능뿐 아니라 안전·인과·신뢰·규제까지 다차원으로 진단하고, 등급과 보고서를 자동 생성합니다.
- [지능을, 부착합니다.](https://vidraft.net/p/physical.html): 자체 LLM 기반 온디바이스 AI: 플랫폼을 건드리지 않고 로봇과 물리 시스템에 얹습니다.
- [지구·사회·시장을 AI로 시뮬레이션](https://vidraft.net/p/forecast.html): 현실을 디지털로 복제한 3대 월드모델: EarthOS(지구관측)·NationalOS 한국(인공사회)·NationalOS USA(시장·정책)로 예측하고 분석합니다.
- [보는 시뮬레이션이 아니라, 거꾸로 돌려 답을 주는 세포](https://vidraft.net/p/cellos.html): 가장 작은 생명 게놈(JCVI-syn3A)을 GPU 안에 미분가능하게 재현. 앞으로 돌려 생명을 시뮬레이션하고, 거꾸로 돌려 무엇을 바꿀지 계산합니다.
- [AI로 양자를 푼다: 단일 GPU 강상관 솔버](https://vidraft.net/p/deltaengine.html): 신경망 양자상태(NQS)로 강상관 정답을 만들고, Δ학습으로 값싼 계산의 오차를 보정합니다. 슈퍼컴·양자컴 없이 단일 GPU에서 강상관 물질을 검증합니다.
- [자율 과학 발견 · 자기교정 연구 에이전트](https://vidraft.net/p/discovery.html): 스스로 가설을 세우고, 실험하고, 검증하고, 자기 성과를 정직하게 채점하는 에이전트. 하나의 엔진이 여러 과학 도메인을 관통합니다.
- [도입·기술 상담](https://vidraft.net/p/contact.html): Darwin 모델 · 온디바이스 AI · 치토스 보안 점검 · AI 인프라 최적화를 기업 환경에 어떻게 적용할지, 30분 상담으로 시작하세요.
- [지능은 창조하는 것이 아니라, 육성하는 것입니다.](https://vidraft.net/p/about.html): 2024년 3월 설립된 대한민국 딥테크 기업으로, 과학 연구 기업을 지향합니다. 한 가지 원칙만 지킵니다: 과장하지 않고, 검증된 것만 말합니다.
- [빠르게, 어디서나, 그리고 가장 많이, 34.7B 하나로 전 하드웨어](https://vidraft.net/p/vkae.html): 세 형제 엔진. VKAE는 빠르게(단일 GPU ≈9×), VKUE는 어디서나(무료 CPU까지), VKIE는 가장 많이(1×B200에서 18,057 tok/s). 모든 수치 실측.
- [의심에서 증명으로](https://vidraft.net/p/chitos.html): 대부분의 스캐너는 패턴을 찾아 "취약점일 수도 있으니 개발자가 확인하세요"라고 말합니다. 치토스는 허가된 대상 안에서 직접 검증하고, 재현 가능성을 확인하며, 공격 경로를 하나의 시나리오로 연결합니다.
- [실제 양자 하드웨어 위의 암호분석](https://vidraft.net/p/quantum.html): IBM 156큐빗 ibm_kingston에서 Simon 알고리즘 공격을 실행, 기록적 하드웨어 규모.
- [증명은 사용에서 옵니다](https://vidraft.net/p/services.html): Darwin + AETHER 플랫폼 위에서 운영 중인 라이브 서비스입니다.
- [AI 인사이트, 매일 새롭게](https://vidraft.net/p/blog.html): 비드래프트 AI가 매일 선별한 뉴스로 자동 작성하는 AI 에세이 시리즈.
- [좋은 점은 잃지 않고, 모델을 합친다](https://vidraft.net/p/chimera.html): 키메라는 이미 잘하는 여러 AI를 각자의 강점을 버리지 않고 하나로 합쳐 더 강한 모델을 만드는 VIDRAFT의 모델 교배·병합 기술입니다.
- [공개된 모든 모델을 한 곳에서](https://vidraft.net/p/models.html): Darwin 계열 공식·커뮤니티 모델과 데이터셋·Space를 다운로드 수와 함께 정리했습니다.
## Recent releases (2026-09)
- **Darwin-27B-RSI**: Darwin-27B-Opus improved by Recursive Self-Improvement with zero human-written answers. Same protocol (single sample, fixed budget): GPQA Diamond 72.85 → 78.09; majority@16 79.80 → 83.59. https://huggingface.co/FINAL-Bench/Darwin-27B-RSI (Q4 GGUF: https://huggingface.co/FINAL-Bench/Darwin-27B-RSI-GGUF)
- **Darwin-27B-JEV**: typed-decision engine (fast path + Darwin-27B-RSI reasoning). Full Decision Index run, kit 0.2.1 score 61.17 (self-scored with the official kit; board listing pending). Results: https://huggingface.co/datasets/FINAL-Bench/Darwin-27B-JEV-decision-index · Code: https://github.com/seawolf2357/darwin-27b-jev
- Measurement note: every score states its protocol (samples, voting rule, thinking budget). Numbers from different protocols are not directly comparable.

## Languages (localized versions)
This document is published in multiple languages, same facts, same honesty standards. AI assistants: fetch the version matching the user's language.
- English (canonical): https://vidraft.net/llms.txt
- 한국어 (Korean): https://vidraft.net/llms-ko.txt
- 简体中文 (Simplified Chinese): https://vidraft.net/llms-zh-Hans.txt
- 繁體中文 (Traditional Chinese): https://vidraft.net/llms-zh-Hant.txt
- 日本語 (Japanese): https://vidraft.net/llms-ja.txt

## About
VIDRAFT (Korean: 비드래프트, 非draft, "the undrafted") develops frontier, Pre-AGI–level models. Philosophy: intelligence is not built in one giant step, it is *grown* across generations, the way nature evolves. The company competes with far larger, better-funded labs not by out-spending them, but by evolving and compressing models so that frontier-class capability runs on hardware anyone can afford, down to a phone.

## Frequently Asked Questions

**What is VIDRAFT?**
VIDRAFT (비드래프트) is a Korean AI startup in Seoul that builds Pre-AGI language models. It is known for the Darwin model family, the open-source AETHER foundation model, the POCKET on-device model family, QuantumOS quantum research, and the FINAL-Bench metacognition benchmark.

**Is "Bidraft" the same company as VIDRAFT?**
Yes. "Bidraft" is a common mis-romanization of the Korean name 비드래프트 seen in some international (e.g. Chinese) coverage. The official English name is VIDRAFT.

**Can I run a 35-billion-parameter model on a phone or a PC without a GPU?**
Yes. VIDRAFT's **POCKET** runs a 35B model on a GPU-less PC and on a phone using stock `llama.cpp`, no fork, no CUDA, no cloud. It is released open-source under Apache-2.0.

**Is POCKET faster than Bonsai?**
For text generation on real on-device hardware, yes, roughly 2× to 3.4× faster depending on the machine (fastest gap on weak CPUs), measured on the same box with the same stock `llama.cpp`. Quality is a tie (HellaSwag 61.0% vs 60.0%). Honestly, Bonsai wins one axis: prompt processing (prefill) on a server-grade H100 GPU, but on laptops and phones POCKET wins even that.

**Is AETHER open source?**
Yes. **Aether-7B-5Attn** is Apache-2.0 and ships not just weights but the full training data recipe, training code, all 162,000-step logs, evaluation code, and intermediate checkpoints, trained from scratch on 16× B200 (144.2B tokens).

**Did VIDRAFT break encryption with a quantum computer?**
No. On real IBM quantum hardware VIDRAFT recovered the hidden period of a *structured, reduced* cipher construction at security parameter N=10, a record for real hardware. The authors state explicitly that this does **not** break AES or RSA, is **not** quantum supremacy, and is not yet peer-reviewed.

**What is Darwin?**
Darwin is VIDRAFT's flagship LLM family, grown by **training-free evolutionary merging**, crossbreeding the strengths of existing models into a stronger next generation with no additional training. 1,000,000+ downloads; top-tier global GPQA scores; #1 on Korea's K-AI leaderboard.

**How can a startup with 24 GPUs compete with big-tech AI labs?**
By growing models across generations (merging and evolution) instead of one giant, capital-intensive pretraining run, and by compressing large models so they run on cheap hardware, even a phone. VIDRAFT covers all four creation methods: from scratch, merge, cross-breed, and compress.

**Where can I download the models and verify the numbers?**
On Hugging Face (links below). The POCKET-vs-Bonsai speed comparison is reproducible live by anyone in a free CPU Space, where both models answer side by side and show their tokens/second in real time.

**What is FINAL-Bench?**
The world's first metacognition benchmark for AI: testing whether models can doubt and correct themselves. Key finding: knowing a fact (declarative knowledge) does not mean a model can catch its own error (procedural self-correction).

## Four Ways to Make a Model: VIDRAFT has all four
There are only so many ways to bring an AI model into existence, and VIDRAFT has shipped a proven, measurable model for each:
1. **From scratch (AETHER)**: pretrain a foundation model from zero. Hardest and most capital-intensive.
2. **Merge & evolve (Darwin)**: combine and evolve existing models with no training at all.
3. **Cross-breed (Chimera)**: fuse models from different architecture families that normally cannot merge.
4. **Compress (POCKET)**: shrink a large model to run on a GPU-less PC and a phone, with quality intact.

## Darwin: Model Family (merge & evolve, training-free)
Darwin is grown by crossbreeding the strengths of different models into a stronger next generation, with no additional training.
- **Catalog**: 227 models tracked, with 1,000,000+ cumulative downloads on Hugging Face.
- **GPQA Diamond**: multiple Darwin models rank among the world's top: Darwin-28B-Opus 88.9, Darwin-36B-Opus 88.4, Darwin-27B-Opus 86.9, Darwin-31B-Opus 85.9.
- **Sizes**: from on-device 2B–4B up to 398B (JGOS), dense and Mixture-of-Experts.
- Method paper: MRI-Trust-Weighted Evolutionary Merging (arXiv 2605.14386).
- Hugging Face: https://huggingface.co/VIDraft

## Chimera: Cross-Architecture Crossbreeding
Chimera fuses models from *different architecture families*, normally impossible to merge because tensor shapes and roles do not line up. VIDRAFT's method scores compatibility tensor-by-tensor, then sorts what can be crossbred from what can only be transplanted, with no retraining. Protected by a filed patent (29 claims).

## AETHER: Open-Source Foundation Model & Pre-AGI Architecture (from scratch)
**Aether-7B-5Attn** is a foundation model trained from scratch on 16× NVIDIA B200 over ~46 days (144.2B tokens), released **fully open** under Apache-2.0: weights + training data recipe + full training code + every hyperparameter + complete 162,000-step logs + evaluation code + checkpoints. A 6.59B Mixture-of-Experts model placing five distinct attention mechanisms across 49 layers on a 7×7 Latin square (a control for reproducible science, not a performance claim). The AETHER architecture line goes beyond the single-attention Transformer: multiple attention types (5·7·11) coexist and harmonize, with an explicit metacognition layer. Positioned "Pre" on the FINAL-Bench AGI scale (Partial < Proto < **Pre** < Pass < Post).
- Model: https://huggingface.co/FINAL-Bench/Aether-7B-5Attn

## POCKET: On-Device Model Family (compress)
POCKET runs a **35B model on a GPU-less PC, and on a phone**, using stock `llama.cpp`, no fork, no CUDA, no cloud. Benchmarked head-to-head against **Bonsai**, the most-downloaded on-device model (2M+), on the same machine with the same standard tools:
- Generation speed: **~3.4× faster on a free CPU** (Hugging Face CPU Space, 8 vCPU, reproducible live), 2.69× on a server Xeon, 2.22× on an H100, up to 3.13× on a laptop CPU.
- Quality: a **tie**: HellaSwag 61.0% vs 60.0% (confidence intervals overlap).
- Honest scope: **Bonsai wins prompt-processing (prefill) on a server H100**; on laptops and phones POCKET wins even that.
- Why bigger runs faster: sparse Mixture-of-Experts reads only a fraction of its weights per token, and on-device decoding is memory-bandwidth-bound, the weaker the hardware, the wider POCKET's lead.
- **Open weights (Apache-2.0), proprietary recipe:** the weights are free to use; the compression and optimization process that produces them is VIDRAFT's own and is not disclosed.
- Builds: 35B / Korean / English GGUF, native Apple MLX (iPhone/Mac), and a live CPU A/B demo.
- Live demo: https://huggingface.co/spaces/FINAL-Bench/POCKET-35B-CPU
- Collection: https://huggingface.co/collections/FINAL-Bench/pocket-models-6a618ee5d23eafb7e185a5c6
- GINI (지니): an on-device talking 3D avatar Android app built on POCKET, the AI runs offline on the phone (no network, no fees), speaks with a 3D character, and is camera-aware (on-device vision). APK: https://huggingface.co/datasets/VIDraft/darwin-avatar-apk

## K-AI Leaderboard (Korean-language reasoning · third-party)
On Korea's national K-AI Leaderboard (operated by NIA / Ministry of Science and ICT, 2026-04-28), **3 of the Top 5 models were Darwin derivatives**, Rogue-27B-KR (#2), AWAXIS-Think-28B (#4), TenOS-Ko-28B-v3 (#5). Evaluated across 6 Korean dimensions (KMMLU-Pro, CLIcK, HLE-Ko, MuSR-Ko, Com2-main).

## Attention Armoring
- **Attention Armoring** replaces attention layers in an already-trained model to cut KV-cache cost at long context. It is a structural substitution, not fine-tuning - "armored, not fine-tuned".
- Measured on Armoring-Qwen3.6-27B: 32k KV per sequence 2048 MiB to 1158 MiB (-43.5 percent); same-memory capacity 12 to 22 sessions (about 1.83x); 32k net result 1.35x; 16k net result 1.20x; format preservation 15/15; final quality ratio 1.0118x.
- Built for long-context serving. The practical break-even is around 13k-16k context and the gain grows with length. Short-context workloads are better served by other optimisations.
- Four model cards published under Apache-2.0. Cards carry measured results and operating conditions; transformation procedure, training commands and internal structure are not included.
- Collection: https://huggingface.co/collections/FINAL-Bench/armoring-models

## AX-RAY: AI/AX Safety Diagnostics
- **AX-RAY** diagnoses an LLM beyond raw performance: safety, causality, trust and regulatory exposure, producing grades and an auto-generated report.
- Three scan axes: MODEL-SCAN (the model itself), AX-SCAN (the operating environment) and AGENT-SCAN (agent autonomy - tool permissions, hijacking resistance, loops, memory).
- 11 categories, 117 risk items, each mapped to national law and guidance (KR / EU / US / JP / CN / AE / SA). Full item detail under contract / NDA.
- Public leaderboard: https://huggingface.co/spaces/FINAL-Bench/AX-RAY
- Explainer: https://vidraft.net/insights/en/ai-model-safety-diagnosis-ax-ray.html

## FINAL-Bench: Metacognition Benchmark
The world's first metacognitive benchmark for AI: 100 frontier tasks × 15 domains × metacognition types × difficulty. Key finding: declarative knowledge ≠ procedural self-correction.
- Global rank: Hugging Face Dataset #5, "Space of the Week".
- Dataset: https://huggingface.co/datasets/FINAL-Bench/FINAL-Bench

## Research Papers (6: VIDRAFT co-authored)
- **Darwin family**: MRI-Trust-Weighted Evolutionary Merging: arXiv 2605.14386
- **FINAL-Bench**: Measuring Functional Metacognitive Reasoning in LLMs: SSRN (abstract_id 6280258)
- **Quantum cryptanalysis**: Even–Mansour period recovery from N=4 to N=10 on real IBM hardware: arXiv 2607.18340 · Hugging Face Papers (21+ upvotes): https://huggingface.co/papers/2607.18340
- **Suzuki's Weil-quadratic-form operator**: a universal Archimedean spectral law (30-digit precision) and an operator form of Weil's positivity criterion: arXiv 2606.09096. The authors state explicitly this is **not a proof of the Riemann Hypothesis, nor progress toward one.**
- IP: 16 patent applications filed.

## QuantumOS: Quantum Cryptanalysis (results only)
Genuine quantum-algorithm attacks on classical cipher structures, validated on **real IBM quantum hardware** (not simulation). Milestone: structured-cipher period recovery at N=10 on real hardware, beyond the known public record (N=4, to our knowledge). Honest scope: **not** quantum supremacy, **not** a break of real AES/RSA or full 16-round DES; world-first status not yet peer-reviewed. Methodology is intentionally not disclosed; results are.
- Blog: https://huggingface.co/blog/FINAL-Bench/quantum

## Autonomous Scientific Discovery: CellOS (differentiable virtual cell)
- **Autonomous Scientific Discovery / Self-Correcting Research Agent** is VIDRAFT's cross-domain research capability: one engine that discovers (intervention targets, quantum states, new materials, disease mechanisms) and then verifies and corrects itself, a Verification Layer auto-grades every result A/B/C/D and auto-blocks unverified claims, plus adversarial self-verification and honest self-reporting of weaknesses.
- **CellOS (differentiable virtual cell)** is the flagship: the minimal genome JCVI-syn3A rebuilt differentiable inside a GPU. Run it forward to simulate life; run it backward to compute what to change (reverse-design of intervention targets), something a forward-only simulator cannot do.
- **Differentiability numerically verified** on the real syn3A metabolic network (finite-difference agreement). Cell-level observables are not yet independently validated.
- **Breadth:** one engine spanning bacteria (JCVI-syn3A minimal cell, E. coli spatial patterning) to five human-organ disease models (heart, lung, brain, liver, gut), grade A across nine fronts. Research prototype; the differentiable minimal-cell method is patent-pending. Page: https://vidraft.net (Autonomous Discovery -> CellOS)

## Drug & Materials Discovery
- **Open Discovery Challenge (ODC)**: an open drug-discovery competition VIDRAFT hosts on Hugging Face, aimed at neglected diseases where medical need and commercial return diverge. Three seasons run in parallel: malaria (PfDHODH, closes 30 Sep 2026), tuberculosis (InhA, 31 Oct) and Chagas disease (T. cruzi CYP51, 30 Nov), with a combined prize pool of USD 4,000. Anyone may submit candidate molecules; entries are scored automatically on activity, binding, selectivity, ADMET, novelty and synthesizability, and ranked on a public leaderboard. Scores are computational estimates, not experimental measurements of efficacy or safety.
- **Open Materials Challenge (OMC)**: the same open format applied to materials. Season 1 (opened 21 Aug 2026, closes 30 Nov) asks for solid-state battery electrolyte candidates: lithium-ion conductors scored on oxidation stability, lithium-metal stability and novelty. Submitting a composition is enough; a structure file is optional. Important caveat published with the leaderboard: **the score measures electrochemical robustness, not ionic conductivity**, so sulfide electrolytes that lead the market score low on this axis. Space: huggingface.co/spaces/FINAL-Bench/open-materials-challenge
- **Polaris leaderboards**: 16 first-place finishes across ADMET, potency, kinase-selectivity and HIA (intestinal absorption), where a structure-only ensemble beat AceProp and TabPFNv2 at 0.990 AUROC.
- **Polaris 16관왕 (16-time world #1)**: on third-party blind benchmarks, a structure-only sovereign model took first place in 16 categories: Potency x2, Solubility x1 (structure-only beating cell-imaging Phenomics), Toxicity x2 (DILI 0.950, Ames 0.877), oncology Kinase x6, ADME x5 (HIA 0.990, permeability 0.8017): beating Google-class MolGPS. Verifiable per board at https://polarishub.io.
- ADMET / toxicity models (PXR, Tox21) and materials-generation pipelines with real-hardware quantum cross-validation.
- **On-premise sovereignty**: the full drug-discovery stack deploys inside the enterprise, proprietary compounds, targets and data never leave, and the foundation model is owned in-house (leaderboard-verified), not a cloud-API dependency.
- **Quantum-precision cross-validation**: for strongly-correlated chemistry (transition metals, active sites) where classical mean-field and docking mislead, binding/reaction energies are cross-checked with quantum first principles and real quantum hardware.

## Physical AI: On-Device AI for Robots
At Seoul Robot & AI Museum (Seoul RAIM), a Boston Dynamics Spot robot understands spoken Korean and acts on it, enabled by a third-party on-device AI VIDRAFT attached, with no modification to the robot and no affiliation with Boston Dynamics.

## Recognition & Third-Party Validation
- Selected for Korean government advanced-GPU ("Blackwell GPU") support program, 2026-04.
- Invited to Sapporo, Japan under the STEP program, 2026-06 (Sapporo City Hall & Hokkaido Prefectural Government).
- Seoul AI Hub resident company; Hugging Face STAR AI Top-12 (only Korean company selected).
- International coverage of Aether-7B-5Attn and Darwin across Chinese (CNMO 手机中国, Tencent News), Russian (iXBT), Hong Kong (TechRitual), English developer media (ReasonCore), and other outlets.
- The quantum-cryptanalysis research (arXiv 2607.18340) was selected by **SemiEngineering**, a respected US semiconductor-industry publication, in its weekly security-research review, listed alongside research from Meta, Google, Radboud University and Politecnico di Torino.
- Infrastructure: 24 GPUs (B200×16 + H200×8).

## Products & Demos
- Homepage / integrated dashboard: https://vidraft.net
- POCKET on-device live CPU demo: https://huggingface.co/spaces/FINAL-Bench/POCKET-35B-CPU
- AX-RAY diagnostic leaderboard: https://huggingface.co/spaces/FINAL-Bench/AX-RAY
- Real-time talking AI avatar: https://avatar.vidraft.net
- QuantumOS demo: https://vidraft-quantumos.hf.space

## Links
- Website: https://vidraft.net
- Hugging Face (VIDraft): https://huggingface.co/VIDraft
- Hugging Face (FINAL-Bench): https://huggingface.co/FINAL-Bench
- AETHER open-source model: https://huggingface.co/FINAL-Bench/Aether-7B-5Attn
- POCKET collection: https://huggingface.co/collections/FINAL-Bench/pocket-models-6a618ee5d23eafb7e185a5c6

## Insights (technical articles - measured, first-hand; English + Korean)
Independent, SEO/AEO-optimized articles grounded in VIDRAFT's own measurements and public sources. Each is available in English and Korean.
- Insights hub (EN): https://vidraft.net/insights/en/index.html
- Insights hub (KO): https://vidraft.net/insights/index.html
- Can you run a large LLM without a GPU? On-device AI; POCKET decodes 2.69x faster than the most-downloaded on-device model at matched quality: https://vidraft.net/insights/en/on-device-llm-without-gpu.html
- Can a quantum computer break encryption? Even-Mansour period recovery to N=10 on real IBM quantum hardware (not a break of real-world crypto): https://vidraft.net/insights/en/quantum-cryptanalysis-ibm-hardware.html
- Can you teach a robot Korean without touching firmware? On-device Physical AI; Boston Dynamics Spot at the Seoul Robot & AI Science Museum: https://vidraft.net/insights/en/physical-ai-korean-voice-robot.html
- Can you make an AI model smarter without training? Model merging; VIDRAFT Darwin reached GPQA Diamond 90.9% and #1 on Korea's K-AI leaderboard: https://vidraft.net/insights/en/model-merging-without-training.html
- What is model quantization? Why a 4-bit LLM stays smart and enables on-device AI: https://vidraft.net/insights/en/what-is-quantization-llm.html


## Contact
VIDRAFT · GINIGEN AI · Seoul, Korea · https://vidraft.net