제미나이 4 아르곤부터 디즈니의 캐릭터AI 제안까지

본문의 각주 번호는 아래 항목과 1:1 로 대응합니다. 정량 주장은 가능한 경우 둘 이상의 출처로 대조했고, 인용문은 원문에서 직접 확인한 것만 실었습니다.

1. 캐릭터AI(Character.ai)의 운영사 Character Technologies 는 2026년 10월 12일 미국 델라웨어 연방파산법원에 자발적 Chapter 11 을 신청했다. 목적은 회사 자산 전부 또는 대부분의 법원 감독 매각이며, 사유로 높은 비용, 소송 책임 가능성, 규제 대응을 들었다Character Technologies Inc. (“Character.ai” or the “Company”), a pioneering artificial intelligence entertainment platform, today announced that it has commenced a voluntary Chapter 11 case in the U.S. Bankruptcy Court for the District of Delaware (the “Court”) to facilitate a value-maximizing sale of all or substantially all of its assetshttps://www.webull.com/news/15738882216961024확인 2026-10-19Character.ai, the startup behind one of the largest artificial intelligence companion apps, filed for Chapter 11 bankruptcy protection Monday, according to a company release and court filings.https://cnbc.com/2026/10/13/character-ai-chapter-11-disney-bid.html확인 2026-10-19These high costs, coupled with potential litigation related liabilities and regulatory engagement have contributed to the Debtor’s inability to generate sufficient cash flows to support its business in the present form.https://cnbc.com/2026/10/13/character-ai-chapter-11-disney-bid.html확인 2026-10-19
   ↩
2. 2026년 10월 19일 KB국민은행 고시 미국 달러 매매기준율은 217회차(12시 15분 54초) 1달러당 1,332.90원이었다217 12:15:54 1,332.90https://obank.kbstar.com/quics?page=C101422&monyCd=USD확인 2026-10-19
   ↩
3. 디즈니는 신청 전 계약으로 캐릭터AI 매각의 스토킹 호스 입찰자가 됐다. 대상은 회사 지식재산 일부와 회사 청구권에 대한 비독점 영구 라이선스이고 값은 2,000만 달러이며, 363조 매각에서 더 높거나 나은 제안이 나오면 밀릴 수 있다Disney will serve as stalking horse bidder for a non-exclusive perpetual license of certain Company intellectual property and certain Company claims, in exchange for $20 million and Disney’s provision of transition services using former Company employees it has hired.https://www.webull.com/news/15738882216961024확인 2026-10-19The transaction is subject to higher or better offers under section 363 of the Bankruptcy Code.https://www.webull.com/news/15738882216961024확인 2026-10-19Disney has agreed to serve as a stalking horse bidder during the sale process to license some of Character.ai’s intellectual property for $20 million.https://cnbc.com/2026/10/13/character-ai-chapter-11-disney-bid.html확인 2026-10-19Disney is willing to pay $20 million for “a non-exclusive perpetual license” of certain Character Technologies intellectual propertyhttps://arstechnica.com/tech-policy/2026/10/disney-offers-20m-for-character-ais-scandal-plagued-chatbot-tech확인 2026-10-19
   ↩
4. 캐릭터AI 서비스는 매각 기간 동안 계속 운영되고, 플랫폼을 이어받는 매수자가 없으면 폐쇄(wound down)되며, 매각 종결 목표는 2026년 11월 말~12월 초다To the extent the Sale Process does not result in a purchase that provides for a continuation of the Character.ai platform, the platform will be wound down.https://www.webull.com/news/15738882216961024확인 2026-10-19the Company anticipates that the completion of the Sale Process and the closing of any resulting transactions will occur in late November or early December 2026.https://www.webull.com/news/15738882216961024확인 2026-10-19the platform could end up being “wound down,”https://cnbc.com/2026/10/13/character-ai-chapter-11-disney-bid.html확인 2026-10-19Any ongoing subscriptions will continue to be charged unless services change to alter existing agreements, the blog said.https://arstechnica.com/tech-policy/2026/10/disney-offers-20m-for-character-ais-scandal-plagued-chatbot-tech확인 2026-10-19
   ↩
5. 구글은 2026년 10월 13일 제미나이 4 아르곤을 제미나이 앱에 내놨고, 구글 AI 프로·울트라 구독자만 쓸 수 있으며 무료 이용자는 대상이 아니다Oct 13, 2026 Gemini 4 Argon is now available in the Gemini apphttps://blog.google/innovation-and-ai/products/gemini-app/gemini-4-argon-availability확인 2026-10-19it's only going to be available for those on AI Pro ($20/mo) and AI Ultra (starting at $99/mo) plans.https://arstechnica.com/ai/2026/10/gemini-4-argon-begins-rolling-out-in-gemini-app-and-antigravity-but-only-if-you-pay확인 2026-10-19
   ↩
6. 오픈AI는 2026년 10월 16일(현지) 챗GPT 의 GPT-6 아스트라 프로에 여러 에이전트의 병렬 실행을 더했다고 릴리스 노트와 X 게시물로 밝혔다. 새 모델이 아니라 기능 개선이고, 병렬 에이전트는 자격이 되는 프로 이용자에게 먼저 풀리며 개수 한도·사용 한도 소모 기준·속도 수치는 공개되지 않았다Updates to GPT-6 Pro model in ChatGPT GPT-6 Pro in the chat tab now offers: Faster responses Improved use of plugins and tools to carry out tasks. Multiple agents that can work on different parts of a task in parallel. You can follow their progress in the conversation and open completed tasks to see what each agent did.https://help.openai.com/en/articles/6825453-chatgpt-release-notes확인 2026-10-19This update is rolling out to eligible Pro users in ChatGPT. Availability in managed workspaces depends on workspace settings and rollout eligibility.https://help.openai.com/en/articles/6825453-chatgpt-release-notes확인 2026-10-19이번 업데이트는 새로운 AI 모델을 공개한 것이 아니라 기존 GPT-6 아스트라 프로의 기능을 개선한 것이다.https://aitimes.com/news/articleView.html?idxno=216375확인 2026-10-19오픈AI는 병렬 작업이 요금제별 사용량에 어떻게 반영되는지, 여러 에이전트를 실행한 요청이 사용 한도를 얼마나 소모하는지에 관한 구체적인 기준을 공개하지 않았다.https://aitimes.com/news/articleView.html?idxno=216375확인 2026-10-19
   ↩
7. 「미공개 물리 문제 98%」는 오픈AI 가 아니라 논문 연구진이 2026년 10월 9일 아카이브에 제출한 논문 TheoPhysEval 의 결과로, GPT-6-Astra 를 코덱스 위에서 돌린 에이전트가 최종 답을 98% 맞혔고 문제당 평균 비용은 3달러 미만이며, 웹 검색을 막아도 정확도가 거의 안 변해 문헌 암기 가능성이 있다We present a suite of 50 unpublished research-level theoretical physics problems with verifiable answers. Without AI assistance, a typical problem in our collection could be solved by a world expert in about a day or less but would take considerably longer for experts in adjacent specialties. We test the ability of off-the-shelf AI agents to tackle these problems autonomously, without human hints. When given access to tools available in a realistic setting, including the ability to search the literature and write scripts for numerical and symbolic work, the best-performing agent, GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an average cost under USD 3 per problem.https://arxiv.org/abs/2610.16733확인 2026-10-19On average, agents show only small changes in accuracy when denied web search and access to background research papers, suggesting significant memorization of the literature.https://arxiv.org/abs/2610.16733확인 2026-10-19[Submitted on 9 Oct 2026]https://arxiv.org/abs/2610.16733확인 2026-10-19
   ↩
8. 앤트로픽은 캐나다 노바스코샤 포인트터퍼에서 에버윈드(EverWind Fuels)와 20년 데이터센터 임대 계약을 맺었다(캐나다 첫 계약). 초기 95MW 에 이어 2028년 중반 325MW 를 늘려 총 420MW 가 되며, 임대료는 비공개다EverWind Fuels announced Friday that U.S. AI company Anthropic has signed a 20-year lease for a 18,600-square metre data centre in Point Tupper, N.S., that would draw 95 megawatts of electricity during its first phase.https://www.cp24.com/news/canada/2026/10/16/everwind-says-10b-ai-data-centre-proposal-largest-private-investment-in-ns-history/확인 2026-10-19A 325-megawatt expansion is expected in mid-2028, which would bring the development’s total energy demand to 420 megawattshttps://www.cp24.com/news/canada/2026/10/16/everwind-says-10b-ai-data-centre-proposal-largest-private-investment-in-ns-history/확인 2026-10-19EverWind officials declined to disclose the value of the leasehttps://www.cp24.com/news/canada/2026/10/16/everwind-says-10b-ai-data-centre-proposal-largest-private-investment-in-ns-history/확인 2026-10-19Oct 16 (Reuters) – Anthropic has signed a lease with EverWind for data center capacity at Point Tupper, Nova Scotia, marking the Claude maker’s first data center agreement in Canadahttps://thedailyguardian.com/news/science-tech/anthropic-signs-lease-for-first-canadian-data-center-766710/확인 2026-10-19
   ↩
9. 리퀴드 AI 는 2026년 10월 13일 LFM2.5-24B-A2B 를 냈다. 전체 24B · 토큰당 활성 약 2B(모델 카드는 2.45B)의 추론 모델이고 문맥은 131,072 토큰이며, 기본판과 후속학습판의 가중치를 공개했고 라이선스는 리퀴드 자체 lfm1.0 이다OCT 13, 2026 Today, we release LFM2.5-24B-A2B, a 24B-parameter reasoning model with just 2B active parameters, built for fast, private agents on-device.https://liquid.ai/blog/lfm2-5-24b-a2b확인 2026-10-19The base (LFM2.5-24B-A2B-Base) and post-trained (LFM2.5-24B-A2B) models are open-weight and available today on Hugging Face and our Playground.https://liquid.ai/blog/lfm2-5-24b-a2b확인 2026-10-19License: lfm1.0https://huggingface.co/LiquidAI/LFM2.5-24B-A2B확인 2026-10-19Active parameters: 2.45Bhttps://huggingface.co/LiquidAI/LFM2.5-24B-A2B확인 2026-10-19Context length: 131,072 tokenshttps://huggingface.co/LiquidAI/LFM2.5-24B-A2B확인 2026-10-19sous licence lfm1.0https://aiweekly.co/fr/alerts/liquid-ai-publie-lfm25-24b-a2b-moe-24-b-avec-245-b-actifs-184-toks-sur-mac-m5확인 2026-10-19
   ↩
10. LFM2.5-24B-A2B 의 에이전트 벤치마크 점수(LiveClaw 48.09 대 38.19)는 리퀴드의 자체 평가이고, 지식·코딩 벤치마크는 더 큰 모델이 앞선다고 리퀴드 스스로 적었으며, 메모리는 애플 M5 맥스에서 15GB 가 든다scoring 48.09 on LiveClaw versus 38.19https://liquid.ai/blog/lfm2-5-24b-a2b확인 2026-10-19Larger models keep an edge on knowledge-heavy benchmarks like GPQA Diamond, HLE, and AA-Omniscience, and on coding benchmarkshttps://liquid.ai/blog/lfm2-5-24b-a2b확인 2026-10-19Apple M5 Max Memory (GB) LFM2.5-24B-A2B 15https://liquid.ai/blog/lfm2-5-24b-a2b확인 2026-10-19
    ↩
11. 엔비디아는 2026년 10월 15일 네모트론 3.5 Super VL(공개 1,200억 파라미터)과 Bolt 74B-A9B 를 냈고, Super VL 은 가중치·데이터셋·훈련 레시피를 공개했으며 라이선스는 엔비디아 NGC 가 OpenMDW License 1.1 로 표기한다October 15, 2026 by Kari Briski 0 Share NVIDIA today released Nemotron 3.5 Super VL, an open 120-billion-parameter model that delivers the fastest task completion and highest accuracy for specialized always-on agents.https://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19Super VL and Bolt 74B-A9B are available now, with Bolt E4B coming Oct. 23.https://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19The model ships with open weights, open datasets and training recipes.https://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19Use of the model is governed by the OpenMDW License Agreement, version 1.1.https://catalog.ngc.nvidia.com/orgs/nim/nvidia/models/nemotron-3.5-super-vl-120b-a12b/-/governing-terms확인 2026-10-19comes with open weights, datasets, and training recipeshttps://www.techzine.eu/news/analytics/145037/nvidia-launches-nemotron-3-5-super-vl-and-bolt-for-ai-agents/확인 2026-10-19
    ↩
12. 엔비디아 Bolt 74B-A9B 는 2비트 양자화로 24GB 메모리에 맞는 로컬 기기용 모델이고, 엔비디아는 같은 급 공개 모델 중 PinchBench 1위·BrowseComp 2위라고 주장한다. 5배 빠르다는 처리량 주장은 Bolt 가 아니라 Super VL 에 대한 것이다(독립 측정 미확인)using a native 2-bit quantization to fit in 24GB of memoryhttps://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19It is the #1 open model in its class on PinchBench, and the #2 open model for BrowseComp.https://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19Nemotron 3.5 Super VL, on Artificial Analysis Intelligence Index leadboard, achieves up to 5x faster throughput compared to other models in its class.https://blogs.nvidia.com/blog/nemotron-3-5-super-vl확인 2026-10-19
    ↩