{"@context":"https://neupai.io/schema/v0.2","@type":"StructuredNewsArticle","identity":{"article_id":"tech42_20260420_moonshot-ai-kimi-k26-release","canonical_url":"https://www.tech42.co.kr/%eb%ac%b8%ec%83%b7ai-%ed%82%a4%eb%af%b8-k2-6-%ec%b6%9c%ec%8b%9c300%ea%b0%9c-ai-%ec%97%90%ec%9d%b4%ec%a0%84%ed%8a%b8-%eb%8f%99%ec%8b%9c-%ec%9a%b4%ec%9a%a9%c2%b7%ec%bd%94%eb%94%a9-%ec%84%b1/?utm_source=rss&utm_medium=rss&utm_campaign=%25eb%25ac%25b8%25ec%2583%25b7ai-%25ed%2582%25a4%25eb%25af%25b8-k2-6-%25ec%25b6%259c%25ec%258b%259c300%25ea%25b0%259c-ai-%25ec%2597%2590%25ec%259d%25b4%25ec%25a0%2584%25ed%258a%25b8-%25eb%258f%2599%25ec%258b%259c-%25ec%259a%25b4%25ec%259a%25a9%25c2%25b7%25ec%25bd%2594%25eb%2594%25a9-%25ec%2584%25b1","ai_url":null,"publisher":{"name":"테크42","domain":"tech42.co.kr","type":"online"},"author":"버트","published_at":"2026-04-20T23:46:00.000Z","updated_at":null,"language":"ko","article_type":"straight_news","originality":"self_produced"},"content":{"headline":"문샷AI '키미 K2.6' 출시…300개 AI 에이전트 동시 운용·코딩 성능 최상위권","summary":"중국 문샷AI가 1조 개 매개변수 규모의 오픈소스 AI 모델 '키미 K2.6'을 공개했다. 코딩 성능 벤치마크에서 GPT와 클로드를 앞섰으며 300개 에이전트 동시 운용이 가능하다.","topics":["AI","인공지능","오픈소스","스타트업"],"geography":["CN"],"entities":[{"name":"문샷AI","canonical_id":"corp:cn:moonshot-ai","type":"company","role_in_article":"primary_subject","metadata":{"ticker":null,"parent":null}},{"name":"키미 K2.6","canonical_id":"product:cn:kimi-k26","type":"product","role_in_article":"primary_subject","metadata":{"ticker":null,"parent":null}},{"name":"GPT","canonical_id":"product:us:gpt","type":"product","role_in_article":"mentioned","metadata":{"ticker":null,"parent":null}},{"name":"클로드 오퍼스","canonical_id":"product:us:claude-opus","type":"product","role_in_article":"mentioned","metadata":{"ticker":null,"parent":null}}],"claims":[{"id":"c1","statement":"중국 AI 스타트업 문샷AI가 최신 오픈소스 AI 모델 '키미 K2.6'을 4월 20일 공개했다","as_of":"2026-04-20","as_of_explicit":true,"as_of_raw":"4월 20일","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c2","statement":"K2.6은 총 1조 개의 매개변수를 갖춘 MoE 아키텍처로 설계됐다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c3","statement":"실제로 동작하는 활성 파라미터는 320억 개다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c4","statement":"SWE-벤치프로에서 58.6점을 기록했다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c5","statement":"GPT 5.4의 57.7점을 앞섰다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":"competitor_comparison","type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c6","statement":"클로드 오퍼스 4.6의 53.4점을 앞섰다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":"competitor_comparison","type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c7","statement":"12시간 이상 4,000번 넘는 도구 호출을 통해 맥 기기에서 AI 모델 추론 속도를 초당 15토큰에서 193토큰으로 대폭 향상시켰다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":"before_after","type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c8","statement":"동시에 300개의 하위 에이전트가 4,000단계를 병렬 처리할 수 있다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c9","statement":"API 이용 요금은 입력 토큰 100만 개당 약 890원(0.60달러)부터 시작한다","as_of":"2026-04-20","as_of_explicit":false,"as_of_raw":"2026년 4월 20일","source_type":"industry_estimate","comparison":null,"type":"estimate","figures":null,"expiry_hint":null,"insight":null}],"ai_emotional_context":{"valence":0,"arousal":0,"primary_emotions":[],"secondary_emotions":[],"emotional_triggers":[]}},"provenance":{"source_chain":["primary_reporting"],"original_source_url":null,"related_articles":[]},"temporal":{"freshness":"recent","next_update_expected":null},"access":{"license":"neupai_standard","attribution_required":true,"structured_data":"free","full_text_available":false,"full_text_access":null}}