https://24-ai.news/hr/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI hr 2026-06-17T00:00:00Z AMD: Instinct MI355X u MLPerf Training v6.0 na 5% zaostatka za NVIDIA-om, 3,5× brži od prošle generacije https://24-ai.news/hr/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI hr 2026-06-17T00:00:00Z Anthropic: Studija o agentskom kodiranju — korisnici donose 70% planskih, Claude 80% izvedbenih odluka https://24-ai.news/hr/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI hr 2026-06-17T00:00:00Z Anthropic: Red Team mapira AI-omogućene kibernapade na MITRE ATT&CK okvir, u suradnji s Verizonom https://24-ai.news/hr/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI hr 2026-06-17T00:00:00Z arXiv:2606.17819: Prvi sistemski benchmark 500 agentskih vještina na 19 konfiguracija modela https://24-ai.news/hr/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI hr 2026-06-17T00:00:00Z arXiv:2606.17930: Rezultati benchmarka ovise o protokolu — inference compute mijenja rang frontier modela https://24-ai.news/hr/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI hr 2026-06-17T00:00:00Z AWS: Novi Bedrock InvokeGuardrailChecks API donosi sigurnosne provjere bez resursa za agentske aplikacije https://24-ai.news/hr/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI hr 2026-06-17T00:00:00Z AWS: P-EAGLE paralelno spekulativno dekodiranje ubrzava inferenciju do 3,97× https://24-ai.news/hr/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI hr 2026-06-17T00:00:00Z AWS: Keširanje kontejnera u SageMaker AI skraćuje skaliranje inferencije do 50% https://24-ai.news/hr/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI hr 2026-06-17T00:00:00Z CNCF: Arhitektonski obrasci za digitalnu suverenost cloud-native platformi pod EU regulativom https://24-ai.news/hr/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI hr 2026-06-17T00:00:00Z Google DeepMind: AI alat za britansko planiranje cilja 50% brže odluke i 1,5 milijuna novih domova https://24-ai.news/hr/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI hr 2026-06-17T00:00:00Z GitHub: GitHub Models zatvoren za nove korisnike — GitHub se povlači iz posluživanja modela https://24-ai.news/hr/news/2026-06-17/langchain-loop-engineering/ 24 AI hr 2026-06-17T00:00:00Z LangChain: 'Loop engineering' — slaganje povratnih petlji oko agenata za pouzdanost i samopoboljšanje https://24-ai.news/hr/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI hr 2026-06-17T00:00:00Z NVIDIA: Blackwell pomeo MLPerf Training 6.0 — najbrži na svih 7 benchmarka, GB300 do 1,6× brži https://24-ai.news/hr/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI hr 2026-06-17T00:00:00Z vLLM: Semantic Router Fusion spaja panel modela uz suca koji sintetizira jedan odgovor https://24-ai.news/en/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI en 2026-06-17T00:00:00Z AMD: Instinct MI355X in MLPerf Training v6.0 Within 5% of NVIDIA, 3.5× Faster Than Previous Generation https://24-ai.news/en/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI en 2026-06-17T00:00:00Z Anthropic: Agentic Coding Study — Users Make 70% of Planning Decisions, Claude 80% of Execution Decisions https://24-ai.news/en/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI en 2026-06-17T00:00:00Z Anthropic: Red Team Maps AI-Enabled Cyberattacks to MITRE ATT&CK Framework, in Partnership with Verizon https://24-ai.news/en/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI en 2026-06-17T00:00:00Z arXiv:2606.17819: First Systematic Benchmark of 500 Agentic Skills Across 19 Model Configurations https://24-ai.news/en/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI en 2026-06-17T00:00:00Z arXiv:2606.17930: Benchmark Results Are Protocol-Dependent — Inference Compute Changes Frontier Model Rankings https://24-ai.news/en/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI en 2026-06-17T00:00:00Z AWS: New Bedrock InvokeGuardrailChecks API Brings Safety Checks Without Resources for Agentic Applications https://24-ai.news/en/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI en 2026-06-17T00:00:00Z AWS: P-EAGLE Parallel Speculative Decoding Accelerates Inference Up to 3.97× https://24-ai.news/en/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI en 2026-06-17T00:00:00Z AWS: Container Caching in SageMaker AI Cuts Inference Scaling Latency by Up to 50% https://24-ai.news/en/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI en 2026-06-17T00:00:00Z CNCF: Architectural Patterns for Digital Sovereignty of Cloud-Native Platforms Under EU Regulation https://24-ai.news/en/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI en 2026-06-17T00:00:00Z Google DeepMind: AI Tool for UK Planning Targets 50% Faster Decisions and 1.5 Million New Homes https://24-ai.news/en/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI en 2026-06-17T00:00:00Z GitHub: GitHub Models Closed to New Customers — GitHub Withdraws from Model Serving https://24-ai.news/en/news/2026-06-17/langchain-loop-engineering/ 24 AI en 2026-06-17T00:00:00Z LangChain: Loop Engineering — Stacking Feedback Loops Around Agents for Reliability and Self-Improvement https://24-ai.news/en/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI en 2026-06-17T00:00:00Z NVIDIA: Blackwell Sweeps MLPerf Training 6.0 — Fastest on All 7 Benchmarks, GB300 Up to 1.6× Faster https://24-ai.news/en/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI en 2026-06-17T00:00:00Z vLLM: Semantic Router Fusion Combines a Model Panel with a Judge That Synthesizes a Single Response https://24-ai.news/de/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI de 2026-06-17T00:00:00Z AMD: Instinct MI355X bei MLPerf Training v6.0 nur 5% hinter NVIDIA, 3,5× schneller als Vorgänger https://24-ai.news/de/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI de 2026-06-17T00:00:00Z Anthropic: Studie zu agentischem Coding — Nutzer treffen 70% der Planungsentscheidungen, Claude 80% der Ausführungsentscheidungen https://24-ai.news/de/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI de 2026-06-17T00:00:00Z Anthropic: Red Team kartiert KI-gestützte Cyberangriffe im MITRE ATT&CK Framework, in Zusammenarbeit mit Verizon https://24-ai.news/de/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI de 2026-06-17T00:00:00Z arXiv:2606.17819: Erster systematischer Benchmark mit 500 agentischen Fähigkeiten auf 19 Modellkonfigurationen https://24-ai.news/de/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI de 2026-06-17T00:00:00Z arXiv:2606.17930: Benchmark-Ergebnisse sind protokollabhängig — Inferenz-Compute verändert das Ranking von Frontier-Modellen https://24-ai.news/de/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI de 2026-06-17T00:00:00Z AWS: Neues Bedrock InvokeGuardrailChecks API bringt Sicherheitsprüfungen ohne Ressourcen für agentische Anwendungen https://24-ai.news/de/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI de 2026-06-17T00:00:00Z AWS: P-EAGLE paralleles spekulatives Dekodieren beschleunigt Inferenz um bis zu 3,97× https://24-ai.news/de/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI de 2026-06-17T00:00:00Z AWS: Container-Caching in SageMaker AI verkürzt Inferenz-Skalierungslatenz um bis zu 50% https://24-ai.news/de/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI de 2026-06-17T00:00:00Z CNCF: Architekturmuster für digitale Souveränität cloud-nativer Plattformen unter EU-Regulierung https://24-ai.news/de/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI de 2026-06-17T00:00:00Z Google DeepMind: KI-Tool für britische Baugenehmigungen zielt auf 50% schnellere Entscheidungen und 1,5 Millionen neue Wohnungen https://24-ai.news/de/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI de 2026-06-17T00:00:00Z GitHub: GitHub Models für Neukunden geschlossen — GitHub zieht sich aus dem Modell-Serving zurück https://24-ai.news/de/news/2026-06-17/langchain-loop-engineering/ 24 AI de 2026-06-17T00:00:00Z LangChain: Loop Engineering — Feedback-Schleifen um Agenten für Zuverlässigkeit und Selbstverbesserung https://24-ai.news/de/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI de 2026-06-17T00:00:00Z NVIDIA: Blackwell dominiert MLPerf Training 6.0 — schnellster auf allen 7 Benchmarks, GB300 bis zu 1,6× schneller https://24-ai.news/de/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI de 2026-06-17T00:00:00Z vLLM: Semantic Router Fusion kombiniert ein Modell-Panel mit einem Richter, der eine einzige Antwort synthetisiert https://24-ai.news/zh/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI zh 2026-06-17T00:00:00Z AMD：Instinct MI355X在MLPerf Training v6.0中与NVIDIA差距缩至5%，性能比上代提升3.5倍 https://24-ai.news/zh/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI zh 2026-06-17T00:00:00Z Anthropic：智能体编码研究 — 用户做出70%规划决策，Claude做出80%执行决策 https://24-ai.news/zh/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI zh 2026-06-17T00:00:00Z Anthropic：红队将AI辅助网络攻击映射至MITRE ATT&CK框架，与Verizon合作完成 https://24-ai.news/zh/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI zh 2026-06-17T00:00:00Z arXiv:2606.17819: 首个针对 19 种模型配置的 500 项智能体技能系统性基准测试 https://24-ai.news/zh/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI zh 2026-06-17T00:00:00Z arXiv:2606.17930: 基准测试结果取决于评估协议——推理计算量改变前沿模型排名 https://24-ai.news/zh/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI zh 2026-06-17T00:00:00Z AWS：新版Bedrock InvokeGuardrailChecks API为智能体应用带来无需资源的安全检查 https://24-ai.news/zh/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI zh 2026-06-17T00:00:00Z AWS：P-EAGLE并行推测解码将推理速度提升最高3.97倍 https://24-ai.news/zh/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI zh 2026-06-17T00:00:00Z AWS：SageMaker AI容器缓存将推理扩展延迟缩短最多50% https://24-ai.news/zh/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI zh 2026-06-17T00:00:00Z CNCF：EU 法规下云原生平台数字主权的架构模式 https://24-ai.news/zh/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI zh 2026-06-17T00:00:00Z Google DeepMind：AI规划工具目标将建筑许可决策提速50%并新建150万套住房 https://24-ai.news/zh/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI zh 2026-06-17T00:00:00Z GitHub：GitHub Models对新用户关闭——GitHub退出模型服务领域 https://24-ai.news/zh/news/2026-06-17/langchain-loop-engineering/ 24 AI zh 2026-06-17T00:00:00Z LangChain「Loop Engineering」——围绕智能体构建反馈循环以实现可靠性与自我改进 https://24-ai.news/zh/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI zh 2026-06-17T00:00:00Z NVIDIA Blackwell横扫MLPerf Training 6.0 — 7项基准全部第一，GB300速度提升最高1.6倍 https://24-ai.news/zh/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI zh 2026-06-17T00:00:00Z vLLM：Semantic Router Fusion将模型面板合并，由裁判模型合成单一回答 https://24-ai.news/ja/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI ja 2026-06-17T00:00:00Z AMD：Instinct MI355X がMLPerf Training v6.0でNVIDIAとの差を5%に縮小、前世代比3.5倍の性能 https://24-ai.news/ja/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI ja 2026-06-17T00:00:00Z Anthropic：エージェントコーディング研究 — ユーザーが計画の70%、Claudeが実行の80%を担う https://24-ai.news/ja/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI ja 2026-06-17T00:00:00Z Anthropic：レッドチームがAI支援サイバー攻撃をMITRE ATT&CKフレームワークにマッピング——Verizonと共同で実施 https://24-ai.news/ja/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI ja 2026-06-17T00:00:00Z arXiv:2606.17819: 19種類のモデル設定で500のエージェンティックスキルを評価する初の体系的ベンチマーク https://24-ai.news/ja/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI ja 2026-06-17T00:00:00Z arXiv:2606.17930: ベンチマーク結果はプロトコルに依存する——推論計算量がフロンティアモデルのランキングを変える https://24-ai.news/ja/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI ja 2026-06-17T00:00:00Z AWS：新しいBedrock InvokeGuardrailChecks APIがエージェント型アプリにリソース不要のセキュリティチェックを提供 https://24-ai.news/ja/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI ja 2026-06-17T00:00:00Z AWS：P-EAGLEの並列スペキュラティブデコーディングで推論を最大3.97倍高速化 https://24-ai.news/ja/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI ja 2026-06-17T00:00:00Z AWS：SageMaker AIのコンテナキャッシングで推論スケーリングのレイテンシを最大50%短縮 https://24-ai.news/ja/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI ja 2026-06-17T00:00:00Z CNCF：EU規制下のクラウドネイティブプラットフォームにおけるデジタル主権のアーキテクチャパターン https://24-ai.news/ja/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI ja 2026-06-17T00:00:00Z Google DeepMind：AIの英国都市計画ツールが建築許可の意思決定を50%高速化し150万戸の住宅新築を目指す https://24-ai.news/ja/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI ja 2026-06-17T00:00:00Z GitHub：GitHub Modelsが新規ユーザーに対して閉鎖——GitHubがモデル提供から撤退 https://24-ai.news/ja/news/2026-06-17/langchain-loop-engineering/ 24 AI ja 2026-06-17T00:00:00Z LangChain「Loop Engineering」——信頼性と自己改善のためにエージェントを囲むフィードバックループの構築 https://24-ai.news/ja/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI ja 2026-06-17T00:00:00Z NVIDIAのBlackwellがMLPerf Training 6.0を制覇 — 全7ベンチマーク首位、GB300は最大1.6倍高速 https://24-ai.news/ja/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI ja 2026-06-17T00:00:00Z vLLM：Semantic Router Fusionがモデルパネルを統合し、審判モデルが単一回答を合成 https://24-ai.news/ko/news/2026-06-17/amd-mi355x-mlperf-training-6/ 24 AI ko 2026-06-17T00:00:00Z AMD：Instinct MI355X가 MLPerf Training v6.0에서 NVIDIA와 5% 차이, 전세대 대비 3.5배 성능 향상 https://24-ai.news/ko/news/2026-06-17/anthropic-claude-code-expertise-study/ 24 AI ko 2026-06-17T00:00:00Z Anthropic：에이전트 코딩 연구 — 사용자가 계획 결정의 70%, Claude가 실행 결정의 80% 담당 https://24-ai.news/ko/news/2026-06-17/anthropic-red-team-attack-navigator/ 24 AI ko 2026-06-17T00:00:00Z Anthropic：레드팀이 AI 지원 사이버 공격을 MITRE ATT&CK 프레임워크에 매핑——Verizon과 협력 https://24-ai.news/ko/news/2026-06-17/arxiv-agentic-skills-benchmark/ 24 AI ko 2026-06-17T00:00:00Z arXiv:2606.17819: 19가지 모델 구성에서 500개 에이전틱 기술을 평가하는 최초의 체계적 벤치마크 https://24-ai.news/ko/news/2026-06-17/arxiv-benchmark-inference-compute/ 24 AI ko 2026-06-17T00:00:00Z arXiv:2606.17930: 벤치마크 결과는 평가 프로토콜에 따라 달라진다——추론 컴퓨팅이 프런티어 모델 순위를 바꾼다 https://24-ai.news/ko/news/2026-06-17/aws-bedrock-guardrail-checks-api/ 24 AI ko 2026-06-17T00:00:00Z AWS：새로운 Bedrock InvokeGuardrailChecks API, 에이전트 앱에 리소스 없이 보안 검사 제공 https://24-ai.news/ko/news/2026-06-17/aws-p-eagle-speculative-decoding/ 24 AI ko 2026-06-17T00:00:00Z AWS：P-EAGLE 병렬 추측 디코딩으로 추론 속도 최대 3.97배 향상 https://24-ai.news/ko/news/2026-06-17/aws-sagemaker-container-caching/ 24 AI ko 2026-06-17T00:00:00Z AWS：SageMaker AI 컨테이너 캐싱으로 추론 스케일링 지연을 최대 50% 단축 https://24-ai.news/ko/news/2026-06-17/cncf-digital-sovereignty-patterns/ 24 AI ko 2026-06-17T00:00:00Z CNCF：EU 규정 하의 클라우드 네이티브 플랫폼을 위한 디지털 주권 아키텍처 패턴 https://24-ai.news/ko/news/2026-06-17/deepmind-uk-planning-ai/ 24 AI ko 2026-06-17T00:00:00Z Google DeepMind：AI 도시계획 도구, 건축 허가 결정 50% 단축 및 150만 가구 신규 주택 목표 https://24-ai.news/ko/news/2026-06-17/github-models-shutdown-new-customers/ 24 AI ko 2026-06-17T00:00:00Z GitHub：GitHub Models 신규 사용자 접근 차단——GitHub, 모델 서비스 사업 철수 https://24-ai.news/ko/news/2026-06-17/langchain-loop-engineering/ 24 AI ko 2026-06-17T00:00:00Z LangChain 『Loop Engineering』——신뢰성과 자기 개선을 위해 에이전트 주변에 피드백 루프 구성하기 https://24-ai.news/ko/news/2026-06-17/nvidia-blackwell-mlperf-training-6/ 24 AI ko 2026-06-17T00:00:00Z NVIDIA Blackwell이 MLPerf Training 6.0 석권 — 7개 벤치마크 전부 1위, GB300 최대 1.6배 빠른 성능 https://24-ai.news/ko/news/2026-06-17/vllm-semantic-router-fusion/ 24 AI ko 2026-06-17T00:00:00Z vLLM：Semantic Router Fusion이 모델 패널을 합쳐 심판 모델이 단일 응답 합성 https://24-ai.news/hr/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI hr 2026-06-16T00:00:00Z AMD: Novi ATOM inference engine za Instinct GPU-e donosi OpenAI-kompatibilan API i MoE optimizacije https://24-ai.news/hr/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI hr 2026-06-16T00:00:00Z Anthropic: Claude Code v2.1.178 uvodi podudaranje parametara alata u dozvolama i ugniježđene skills https://24-ai.news/hr/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI hr 2026-06-16T00:00:00Z Anthropic: Originalni modeli Claude Sonnet 4 i Opus 4 povučeni — migracija na Sonnet 4.6 i Opus 4.8 https://24-ai.news/hr/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI hr 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBench mjeri demografsku diskriminaciju u akcijama LLM agenata https://24-ai.news/hr/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI hr 2026-06-16T00:00:00Z arXiv:2606.17005: Bayesov okvir za reviziju otkriva da ljestvice AI modela kriju nekompatibilne povijesti https://24-ai.news/hr/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI hr 2026-06-16T00:00:00Z AWS: Google Gemma 4 modeli dostupni na Amazon Bedrocku — tri varijante uz Apache 2.0 licencu https://24-ai.news/hr/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI hr 2026-06-16T00:00:00Z AWS: Strands Evals SDK automatizira otkrivanje kvarova AI agenata i analizu korijenskog uzroka https://24-ai.news/hr/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI hr 2026-06-16T00:00:00Z CNCF: Oracleova donacija od 3 milijuna dolara u OCI kreditima ubrzava Arm64 podršku u 12+ projekata https://24-ai.news/hr/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI hr 2026-06-16T00:00:00Z GitHub: Otvoreni Multilingual Repositories Dataset s 80 milijuna redaka i 40 milijuna repozitorija https://24-ai.news/hr/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI hr 2026-06-16T00:00:00Z Google: Gemini API najavljuje gašenje Veo i Imagen 4 modela — Veo 3.0 odlazi već 30. lipnja https://24-ai.news/hr/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI hr 2026-06-16T00:00:00Z LangChain: Fino ugođeni Qwen-3.5-35B kao 'trace judge' 10–100× jeftiniji od frontier modela https://24-ai.news/hr/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI hr 2026-06-16T00:00:00Z Microsoft: Potrošnja energije po AI upitu 4–20× manja od ranijih procjena, moguće 8–20× uštede https://24-ai.news/hr/news/2026-06-16/openai-partner-network-enterprise/ 24 AI hr 2026-06-16T00:00:00Z OpenAI: Pokrenut Partner Network s 150 milijuna dolara za ubrzanje enterprise AI usvajanja https://24-ai.news/en/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI en 2026-06-16T00:00:00Z AMD: New ATOM Inference Engine for Instinct GPUs Brings OpenAI-Compatible API and MoE Optimizations https://24-ai.news/en/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI en 2026-06-16T00:00:00Z Anthropic: Claude Code v2.1.178 Introduces Tool Parameter Matching in Permissions and Nested Skills https://24-ai.news/en/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI en 2026-06-16T00:00:00Z Anthropic: Original Claude Sonnet 4 and Opus 4 Models Retired — Migration to Sonnet 4.6 and Opus 4.8 https://24-ai.news/en/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI en 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBench Measures Demographic Discrimination in LLM Agent Actions https://24-ai.news/en/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI en 2026-06-16T00:00:00Z arXiv:2606.17005: Bayesian Framework for Auditing Reveals That AI Leaderboards Hide Incompatible Histories https://24-ai.news/en/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI en 2026-06-16T00:00:00Z AWS: Google Gemma 4 Models Available on Amazon Bedrock — Three Variants Under Apache 2.0 License https://24-ai.news/en/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI en 2026-06-16T00:00:00Z AWS: Strands Evals SDK Automates AI Agent Failure Detection and Root Cause Analysis https://24-ai.news/en/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI en 2026-06-16T00:00:00Z CNCF: Oracle's $3 Million Donation in OCI Credits Accelerates Arm64 Support in 12+ Projects https://24-ai.news/en/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI en 2026-06-16T00:00:00Z GitHub: Open Multilingual Repositories Dataset with 80 Million Rows and 40 Million Repositories https://24-ai.news/en/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI en 2026-06-16T00:00:00Z Google: Gemini API Announces Shutdown Deadlines for Veo and Imagen 4 Models — Veo 3.0 Leaves June 30 https://24-ai.news/en/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI en 2026-06-16T00:00:00Z LangChain: Fine-Tuned Qwen-3.5-35B as a Trace Judge 10–100× Cheaper Than Frontier Models https://24-ai.news/en/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI en 2026-06-16T00:00:00Z Microsoft: Energy Consumption per AI Query 4–20× Lower Than Earlier Estimates, 8–20× Savings Possible https://24-ai.news/en/news/2026-06-16/openai-partner-network-enterprise/ 24 AI en 2026-06-16T00:00:00Z OpenAI: Partner Network Launched with $150 Million to Accelerate Enterprise AI Adoption https://24-ai.news/de/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI de 2026-06-16T00:00:00Z AMD: Neuer ATOM Inference Engine für Instinct GPUs bietet OpenAI-kompatible API und MoE-Optimierungen https://24-ai.news/de/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI de 2026-06-16T00:00:00Z Anthropic: Claude Code v2.1.178 führt Parameter-Abgleich bei Berechtigungen und verschachtelte Skills ein https://24-ai.news/de/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI de 2026-06-16T00:00:00Z Anthropic: Originale Claude Sonnet 4 und Opus 4 Modelle eingestellt — Migration auf Sonnet 4.6 und Opus 4.8 https://24-ai.news/de/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI de 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBench misst demografische Diskriminierung in Aktionen von LLM-Agenten https://24-ai.news/de/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI de 2026-06-16T00:00:00Z arXiv:2606.17005: Bayesscher Rahmen für Audits zeigt, dass KI-Bestenlisten unvereinbare Historien verbergen https://24-ai.news/de/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI de 2026-06-16T00:00:00Z AWS: Google Gemma 4 Modelle auf Amazon Bedrock verfügbar — drei Varianten unter Apache-2.0-Lizenz https://24-ai.news/de/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI de 2026-06-16T00:00:00Z AWS: Strands Evals SDK automatisiert KI-Agenten-Fehlererkennung und Ursachenanalyse https://24-ai.news/de/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI de 2026-06-16T00:00:00Z CNCF: Oracles Spende von 3 Millionen Dollar in OCI-Credits beschleunigt Arm64-Unterstützung in 12+ Projekten https://24-ai.news/de/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI de 2026-06-16T00:00:00Z GitHub: Offener Multilingual Repositories Dataset mit 80 Millionen Zeilen und 40 Millionen Repositories https://24-ai.news/de/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI de 2026-06-16T00:00:00Z Google: Gemini API kündigt Abschalttermine für Veo und Imagen 4 an — Veo 3.0 verschwindet bereits am 30. Juni https://24-ai.news/de/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI de 2026-06-16T00:00:00Z LangChain: Feinabgestimmter Qwen-3.5-35B als Trace-Judge 10–100× günstiger als Frontier-Modelle https://24-ai.news/de/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI de 2026-06-16T00:00:00Z Microsoft: Energieverbrauch pro KI-Anfrage 4–20× niedriger als frühere Schätzungen, 8–20× Einsparungen möglich https://24-ai.news/de/news/2026-06-16/openai-partner-network-enterprise/ 24 AI de 2026-06-16T00:00:00Z OpenAI: Partner Network mit 150 Millionen Dollar zur Beschleunigung der Enterprise-KI-Einführung gestartet https://24-ai.news/zh/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI zh 2026-06-16T00:00:00Z AMD：面向Instinct GPU的全新ATOM推理引擎提供OpenAI兼容API和MoE优化 https://24-ai.news/zh/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI zh 2026-06-16T00:00:00Z Anthropic：Claude Code v2.1.178引入权限中的工具参数匹配与嵌套技能 https://24-ai.news/zh/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI zh 2026-06-16T00:00:00Z Anthropic：原版Claude Sonnet 4与Opus 4退役——迁移至Sonnet 4.6与Opus 4.8 https://24-ai.news/zh/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI zh 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBench测量LLM代理行为中的人口统计歧视 https://24-ai.news/zh/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI zh 2026-06-16T00:00:00Z arXiv:2606.17005: 贝叶斯审计框架揭示AI排行榜隐藏不兼容的历史记录 https://24-ai.news/zh/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI zh 2026-06-16T00:00:00Z AWS：Google Gemma 4模型登陆Amazon Bedrock——三种变体采用Apache 2.0许可证 https://24-ai.news/zh/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI zh 2026-06-16T00:00:00Z AWS：Strands Evals SDK自动化AI代理故障检测与根因分析 https://24-ai.news/zh/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI zh 2026-06-16T00:00:00Z CNCF：Oracle捐赠300万美元OCI积分，加速12个以上项目的Arm64支持 https://24-ai.news/zh/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI zh 2026-06-16T00:00:00Z GitHub：开放多语言代码库数据集——8000万行记录、4000万个仓库 https://24-ai.news/zh/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI zh 2026-06-16T00:00:00Z Google：Gemini API宣布Veo和Imagen 4模型停用时间表——Veo 3.0将于6月30日退场 https://24-ai.news/zh/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI zh 2026-06-16T00:00:00Z LangChain：微调后的Qwen-3.5-35B作为轨迹评判器，比前沿模型便宜10至100倍 https://24-ai.news/zh/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI zh 2026-06-16T00:00:00Z Microsoft：每次AI查询能耗比早期估算低4至20倍，可实现8至20倍节能 https://24-ai.news/zh/news/2026-06-16/openai-partner-network-enterprise/ 24 AI zh 2026-06-16T00:00:00Z OpenAI：启动合作伙伴网络，投入1.5亿美元加速企业AI采用 https://24-ai.news/ja/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI ja 2026-06-16T00:00:00Z AMD：Instinct GPU向けの新推論エンジンATOMがOpenAI互換APIとMoE最適化を提供 https://24-ai.news/ja/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI ja 2026-06-16T00:00:00Z Anthropic：Claude Code v2.1.178がパーミッションへのツールパラメータマッチングとネストされたスキルを導入 https://24-ai.news/ja/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI ja 2026-06-16T00:00:00Z Anthropic：Claude Sonnet 4とOpus 4の初代モデルが廃止——Sonnet 4.6とOpus 4.8への移行を推奨 https://24-ai.news/ja/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI ja 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBenchがLLMエージェントの行動における人口統計的差別を測定 https://24-ai.news/ja/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI ja 2026-06-16T00:00:00Z arXiv:2606.17005: ベイズ監査フレームワークがAIリーダーボードに非互換な歴史が隠れていることを示す https://24-ai.news/ja/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI ja 2026-06-16T00:00:00Z AWS：Google Gemma 4モデルがAmazon Bedrockに登場——3種類のバリアントがApache 2.0ライセンスで提供 https://24-ai.news/ja/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI ja 2026-06-16T00:00:00Z AWS：Strands Evals SDKがAIエージェントの障害検出と根本原因分析を自動化 https://24-ai.news/ja/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI ja 2026-06-16T00:00:00Z CNCF：Oracleの300万ドルOCIクレジット寄付が12以上のプロジェクトのArm64サポートを加速 https://24-ai.news/ja/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI ja 2026-06-16T00:00:00Z GitHub：8000万行・4000万リポジトリのオープンな多言語リポジトリデータセットを公開 https://24-ai.news/ja/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI ja 2026-06-16T00:00:00Z Google：Gemini APIがVeoとImagen 4モデルの廃止期限を発表——Veo 3.0は6月30日に終了 https://24-ai.news/ja/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI ja 2026-06-16T00:00:00Z LangChain：ファインチューニングされたQwen-3.5-35Bが「トレースジャッジ」としてフロンティアモデルより10〜100倍低コスト https://24-ai.news/ja/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI ja 2026-06-16T00:00:00Z Microsoft：AIクエリ1件あたりのエネルギー消費量は従来の推定より4〜20倍低く、8〜20倍の節約が可能 https://24-ai.news/ja/news/2026-06-16/openai-partner-network-enterprise/ 24 AI ja 2026-06-16T00:00:00Z OpenAI：1億5000万ドルのパートナーネットワークを立ち上げ、エンタープライズAI導入を加速 https://24-ai.news/ko/news/2026-06-16/amd-atom-inference-engine-instinct/ 24 AI ko 2026-06-16T00:00:00Z AMD：Instinct GPU용 신규 ATOM 추론 엔진, OpenAI 호환 API와 MoE 최적화 제공 https://24-ai.news/ko/news/2026-06-16/anthropic-claude-code-2-1-178/ 24 AI ko 2026-06-16T00:00:00Z Anthropic：Claude Code v2.1.178, 권한의 도구 파라미터 매칭과 중첩 스킬 도입 https://24-ai.news/ko/news/2026-06-16/anthropic-sonnet-4-opus-4-retired/ 24 AI ko 2026-06-16T00:00:00Z Anthropic：Claude Sonnet 4·Opus 4 초기 모델 폐기——Sonnet 4.6·Opus 4.8로 마이그레이션 권고 https://24-ai.news/ko/news/2026-06-16/arxiv-agentfairbench-agent-discrimination/ 24 AI ko 2026-06-16T00:00:00Z arXiv:2606.16723: AgentFairBench, LLM 에이전트 행동에서의 인구통계 차별 측정 https://24-ai.news/ko/news/2026-06-16/arxiv-bayesian-audit-ai-leaderboards/ 24 AI ko 2026-06-16T00:00:00Z arXiv:2606.17005: 베이즈 감사 프레임워크, AI 리더보드가 비호환 역사를 숨긴다는 것을 보여줌 https://24-ai.news/ko/news/2026-06-16/aws-bedrock-gemma-4-models/ 24 AI ko 2026-06-16T00:00:00Z AWS：Google Gemma 4 모델, Amazon Bedrock 출시——Apache 2.0 라이선스 세 가지 변형 제공 https://24-ai.news/ko/news/2026-06-16/aws-strands-evals-agent-failure-rca/ 24 AI ko 2026-06-16T00:00:00Z AWS：Strands Evals SDK, AI 에이전트 장애 감지와 근본 원인 분석 자동화 https://24-ai.news/ko/news/2026-06-16/cncf-oracle-oci-arm64-credits/ 24 AI ko 2026-06-16T00:00:00Z CNCF：Oracle의 300만 달러 OCI 크레딧 기부, 12개 이상 프로젝트의 Arm64 지원 가속화 https://24-ai.news/ko/news/2026-06-16/github-multilingual-repositories-dataset/ 24 AI ko 2026-06-16T00:00:00Z GitHub：8000만 행·4000만 저장소 오픈 다국어 저장소 데이터셋 공개 https://24-ai.news/ko/news/2026-06-16/google-veo-imagen4-deprecation-deadlines/ 24 AI ko 2026-06-16T00:00:00Z Google：Gemini API, Veo·Imagen 4 모델 폐기 일정 발표——Veo 3.0은 6월 30일 종료 https://24-ai.news/ko/news/2026-06-16/langchain-trace-judge-qwen-finetune/ 24 AI ko 2026-06-16T00:00:00Z LangChain：파인튜닝된 Qwen-3.5-35B, 프론티어 모델보다 10~100배 저렴한 트레이스 판단자로 활약 https://24-ai.news/ko/news/2026-06-16/microsoft-ai-query-energy-efficiency/ 24 AI ko 2026-06-16T00:00:00Z Microsoft：AI 쿼리당 에너지 소비, 기존 추정치보다 4~20배 낮아——8~20배 절감 가능 https://24-ai.news/ko/news/2026-06-16/openai-partner-network-enterprise/ 24 AI ko 2026-06-16T00:00:00Z OpenAI：1억 5000만 달러 파트너 네트워크 출범, 기업 AI 도입 가속화