https://24-ai.news/hr/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI hr 2026-06-27T00:00:00Z AMD: MXFP4/MXFP6 mixed-precision kvantizacija na MI355X — do 29% veći throughput https://24-ai.news/hr/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI hr 2026-06-27T00:00:00Z Anthropic: API rate limiti podignuti — Sonnet i Haiku sad na razini Opusa, tri tiera https://24-ai.news/hr/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI hr 2026-06-27T00:00:00Z Anthropic: Economic Index otkriva ritmove i obrasce korištenja AI-ja kroz dan https://24-ai.news/hr/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.26649: Upute agentima pretvorene u formalno verificirani policy-as-code https://24-ai.news/hr/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.26836: Benchmarci propuštaju 82% stvarnih sposobnosti AI modela https://24-ai.news/hr/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.27288: Kad kombiniranje LLM-ova stvarno pomaže — co-failure ceiling na 67 frontier modela https://24-ai.news/hr/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.26758: EGG — višeagentni okvir generira GPU kernele 2,13× brže od PyTorcha https://24-ai.news/hr/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard — brza moderacija sadržaja bez chain-of-thought dostiže teške reasonere https://24-ai.news/hr/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI hr 2026-06-27T00:00:00Z arXiv:2606.27009: Semantičko rano zaustavljanje smanjuje trošak agentskih petlji za 38% https://24-ai.news/hr/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI hr 2026-06-27T00:00:00Z AWS: Stripe pokrenuo 100+ AI agenata za financijsku usklađenost — lekcije iz produkcije https://24-ai.news/hr/news/2026-06-27/claude-code-2-1-195/ 24 AI hr 2026-06-27T00:00:00Z Anthropic: Claude Code v2.1.195 — onemogućavanje miša, poboljšan voice dictation i popravci hookova https://24-ai.news/hr/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI hr 2026-06-27T00:00:00Z GitHub: Desktop 3.6 donosi Git worktrees i AI razrješavanje merge konflikata https://24-ai.news/hr/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI hr 2026-06-27T00:00:00Z Google: Gemini Nano na Pixelu 50%+ brži uz zamrznutu multi-token predikciju https://24-ai.news/hr/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI hr 2026-06-27T00:00:00Z LangChain: prompt caching u Deep Agents okviru smanjuje latenciju dugotrajnih agenata https://24-ai.news/hr/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI hr 2026-06-27T00:00:00Z OpenAI: GPT-5.6 Sol najavljen u previewu — kodiranje, znanost i kibersigurnost https://24-ai.news/en/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI en 2026-06-27T00:00:00Z AMD: MXFP4/MXFP6 mixed-precision quantization on MI355X delivers up to 29% higher throughput https://24-ai.news/en/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI en 2026-06-27T00:00:00Z Anthropic: API rate limits raised — Sonnet and Haiku now match Opus across three tiers https://24-ai.news/en/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI en 2026-06-27T00:00:00Z Anthropic: Economic Index reveals daily rhythms and patterns of AI usage https://24-ai.news/en/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.26649: Agent instructions converted into formally verified policy-as-code https://24-ai.news/en/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.26836: Benchmarks miss 82% of AI models' real capabilities https://24-ai.news/en/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.27288: When combining LLMs really helps — co-failure ceiling across 67 frontier models https://24-ai.news/en/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.26758: EGG — multi-agent framework generates GPU kernels 2.13x faster than PyTorch https://24-ai.news/en/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard — fast content moderation without chain-of-thought matches heavy reasoners https://24-ai.news/en/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI en 2026-06-27T00:00:00Z arXiv:2606.27009: Semantic early stopping cuts agent loop token cost by 38% https://24-ai.news/en/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI en 2026-06-27T00:00:00Z AWS: Stripe launched 100+ AI agents for financial compliance — lessons from production https://24-ai.news/en/news/2026-06-27/claude-code-2-1-195/ 24 AI en 2026-06-27T00:00:00Z Anthropic: Claude Code v2.1.195 — mouse disabling, improved voice dictation, and hook fixes https://24-ai.news/en/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI en 2026-06-27T00:00:00Z GitHub: Desktop 3.6 brings Git worktrees and AI merge conflict resolution https://24-ai.news/en/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI en 2026-06-27T00:00:00Z Google: Gemini Nano on Pixel is 50%+ faster with frozen multi-token prediction https://24-ai.news/en/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI en 2026-06-27T00:00:00Z LangChain: prompt caching in Deep Agents framework reduces long-running agent latency https://24-ai.news/en/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI en 2026-06-27T00:00:00Z OpenAI: GPT-5.6 Sol announced in preview — coding, science, and cybersecurity https://24-ai.news/de/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI de 2026-06-27T00:00:00Z AMD: MXFP4/MXFP6 Mixed-Precision-Quantisierung auf MI355X — bis zu 29% mehr Durchsatz https://24-ai.news/de/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI de 2026-06-27T00:00:00Z Anthropic: API-Rate-Limits angehoben — Sonnet und Haiku jetzt auf Opus-Niveau, drei Stufen https://24-ai.news/de/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI de 2026-06-27T00:00:00Z Anthropic: Economic Index zeigt Tagesrhythmen und Nutzungsmuster bei KI https://24-ai.news/de/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.26649: Agentenanweisungen werden zu formal verifiziertem Policy-as-Code https://24-ai.news/de/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.26836: Benchmarks übersehen 82% der tatsächlichen KI-Modellfähigkeiten https://24-ai.news/de/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.27288: Wann die Kombination von Sprachmodellen wirklich hilft — Co-Failure Ceiling bei 67 Frontier-Modellen https://24-ai.news/de/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.26758: EGG — Mehrагenten-Framework generiert GPU-Kernel 2,13× schneller als PyTorch https://24-ai.news/de/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard — schnelle Inhaltsmoderation ohne Chain-of-Thought erreicht schwere Reasoning-Modelle https://24-ai.news/de/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI de 2026-06-27T00:00:00Z arXiv:2606.27009: Semantisches frühzeitiges Stoppen senkt Kosten von Agenten-Schleifen um 38% https://24-ai.news/de/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI de 2026-06-27T00:00:00Z AWS: Stripe hat über 100 KI-Agenten für die Finanzcompliance eingeführt — Erkenntnisse aus dem Produktionsbetrieb https://24-ai.news/de/news/2026-06-27/claude-code-2-1-195/ 24 AI de 2026-06-27T00:00:00Z Anthropic: Claude Code v2.1.195 — Maus deaktivieren, verbessertes Voice Dictation und Hook-Korrekturen https://24-ai.news/de/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI de 2026-06-27T00:00:00Z GitHub: Desktop 3.6 bringt Git Worktrees und KI-gestützte Merge-Konfliktlösung https://24-ai.news/de/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI de 2026-06-27T00:00:00Z Google: Gemini Nano auf Pixel 50%+ schneller durch eingefrorene Multi-Token-Prediktion https://24-ai.news/de/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI de 2026-06-27T00:00:00Z LangChain: Prompt Caching im Deep Agents-Framework senkt Latenz langlebiger Agenten https://24-ai.news/de/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI de 2026-06-27T00:00:00Z OpenAI: GPT-5.6 Sol im Preview angekündigt — Kodierung, Wissenschaft und Cybersicherheit https://24-ai.news/zh/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI zh 2026-06-27T00:00:00Z AMD：MI355X上的MXFP4/MXFP6混合精度量化——吞吐量最高提升29% https://24-ai.news/zh/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI zh 2026-06-27T00:00:00Z Anthropic：API速率限制提升——Sonnet和Haiku与Opus同等，三个层级 https://24-ai.news/zh/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI zh 2026-06-27T00:00:00Z Anthropic经济指数：揭示全天AI使用的节律与模式 https://24-ai.news/zh/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.26649: AI代理指令转换为形式化验证的策略即代码 https://24-ai.news/zh/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.26836: 标准基准遗漏了82%的AI模型真实能力 https://24-ai.news/zh/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.27288: 组合LLM何时真正有效——67个前沿模型的共同失败上限 https://24-ai.news/zh/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.26758: EGG——多代理框架生成GPU内核比PyTorch快2.13倍 https://24-ai.news/zh/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard——无需思维链的快速内容审核媲美重量级推理器 https://24-ai.news/zh/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI zh 2026-06-27T00:00:00Z arXiv:2606.27009: 语义早停将代理循环成本降低38% https://24-ai.news/zh/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI zh 2026-06-27T00:00:00Z AWS：Stripe部署100+个AI代理用于金融合规——来自生产环境的经验教训 https://24-ai.news/zh/news/2026-06-27/claude-code-2-1-195/ 24 AI zh 2026-06-27T00:00:00Z Anthropic：Claude Code v2.1.195——禁用鼠标、改进语音听写和钩子修复 https://24-ai.news/zh/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI zh 2026-06-27T00:00:00Z GitHub：Desktop 3.6引入Git工作树和AI合并冲突解决 https://24-ai.news/zh/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI zh 2026-06-27T00:00:00Z Google：冻结多令牌预测使Pixel上的Gemini Nano推理速度提升50%以上 https://24-ai.news/zh/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI zh 2026-06-27T00:00:00Z LangChain：Deep Agents框架中的提示缓存降低长期代理延迟 https://24-ai.news/zh/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI zh 2026-06-27T00:00:00Z OpenAI：GPT-5.6 Sol预览发布——编码、科学与网络安全 https://24-ai.news/ja/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI ja 2026-06-27T00:00:00Z AMD: MI355XでMXFP4/MXFP6混合精度量子化 — スループット最大29%向上 https://24-ai.news/ja/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI ja 2026-06-27T00:00:00Z Anthropic：APIレート制限を引き上げ — SonnetとHaikuがOpusと同水準に、3段階のティア制 https://24-ai.news/ja/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI ja 2026-06-27T00:00:00Z Anthropic：Economic IndexがAI利用の日常的リズムとパターンを明らかに https://24-ai.news/ja/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.26649: エージェントへの指示を形式検証済みのpolicy-as-codeに変換 https://24-ai.news/ja/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.26836: 標準ベンチマークはAIモデルの実力の82%を見落としている https://24-ai.news/ja/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.27288: LLMアンサンブルが実際に効果的な場合——67のフロンティアモデルでco-failure ceiling検証 https://24-ai.news/ja/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.26758: EGG——マルチエージェントフレームワークがPyTorch比2.13倍高速なGPUカーネルを生成 https://24-ai.news/ja/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard——chain-of-thoughtなしの高速コンテンツモデレーションが重量級推論モデルに匹敵 https://24-ai.news/ja/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI ja 2026-06-27T00:00:00Z arXiv:2606.27009: 意味的早期停止でエージェントループのトークンコストを38%削減 https://24-ai.news/ja/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI ja 2026-06-27T00:00:00Z AWS：Stripeが金融コンプライアンス向けに100以上のAIエージェントを本番稼働——現場からの教訓 https://24-ai.news/ja/news/2026-06-27/claude-code-2-1-195/ 24 AI ja 2026-06-27T00:00:00Z Anthropic：Claude Code v2.1.195——マウス無効化、音声ディクテーション改善、フックの修正 https://24-ai.news/ja/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI ja 2026-06-27T00:00:00Z GitHub：Desktop 3.6がGit worktreesとAIによるマージコンフリクト解決を導入 https://24-ai.news/ja/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI ja 2026-06-27T00:00:00Z Google：凍結マルチトークン予測でPixelのGemini Nanoが50%以上高速化 https://24-ai.news/ja/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI ja 2026-06-27T00:00:00Z LangChain：Deep AgentsフレームワークのプロンプトキャッシングでLLMエージェントの遅延を削減 https://24-ai.news/ja/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI ja 2026-06-27T00:00:00Z OpenAI：GPT-5.6 Solをプレビューで発表——コーディング・科学・サイバーセキュリティ https://24-ai.news/ko/news/2026-06-27/amd-mi355x-mxfp4-quantization/ 24 AI ko 2026-06-27T00:00:00Z AMD: MI355X에서 MXFP4/MXFP6 혼합 정밀도 양자화로 처리량 29% 향상 https://24-ai.news/ko/news/2026-06-27/anthropic-api-rate-limits-three-tiers/ 24 AI ko 2026-06-27T00:00:00Z Anthropic: API 속도 제한 상향 — Sonnet·Haiku가 Opus 수준으로, 3단계 티어 도입 https://24-ai.news/ko/news/2026-06-27/anthropic-economic-index-cadences/ 24 AI ko 2026-06-27T00:00:00Z Anthropic: Economic Index, 하루 중 AI 활용 리듬과 패턴 분석 https://24-ai.news/ko/news/2026-06-27/arxiv-autoformalization-agent-policy-code/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.26649: AI 에이전트 지침을 형식적으로 검증된 정책 코드로 자동 변환 https://24-ai.news/ko/news/2026-06-27/arxiv-capability-frontier-benchmarks/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.26836: 표준 벤치마크, AI 모델 실제 능력의 82%를 놓치고 있다 https://24-ai.news/ko/news/2026-06-27/arxiv-co-failure-ceiling-llm-ensembles/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.27288: LLM 앙상블이 실제로 도움이 될 때 — 67개 프론티어 모델의 공동 실패 천장 https://24-ai.news/ko/news/2026-06-27/arxiv-egg-expert-guided-kernels/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.26758: EGG — 다중 에이전트 프레임워크, PyTorch보다 2.13배 빠른 GPU 커널 생성 https://24-ai.news/ko/news/2026-06-27/arxiv-leanguard-fast-content-moderation/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.26686: LeanGuard — CoT 없이 빠른 콘텐츠 모더레이션으로 무거운 추론 모델과 대등 https://24-ai.news/ko/news/2026-06-27/arxiv-semantic-early-stopping-agents/ 24 AI ko 2026-06-27T00:00:00Z arXiv:2606.27009: 의미론적 조기 중단으로 에이전트 루프 비용 38% 절감 https://24-ai.news/ko/news/2026-06-27/aws-stripe-ai-agents-financial-compliance/ 24 AI ko 2026-06-27T00:00:00Z AWS: Stripe, 금융 규정 준수를 위해 100개 이상의 AI 에이전트 출시 — 프로덕션 교훈 https://24-ai.news/ko/news/2026-06-27/claude-code-2-1-195/ 24 AI ko 2026-06-27T00:00:00Z Anthropic: Claude Code v2.1.195 — 마우스 비활성화, 개선된 음성 받아쓰기, 훅 수정 https://24-ai.news/ko/news/2026-06-27/github-desktop-3-6-worktrees-copilot/ 24 AI ko 2026-06-27T00:00:00Z GitHub: Desktop 3.6, Git 워크트리와 AI 머지 충돌 해결 기능 출시 https://24-ai.news/ko/news/2026-06-27/google-gemini-nano-frozen-mtp-pixel/ 24 AI ko 2026-06-27T00:00:00Z Google: 고정된 멀티 토큰 예측으로 Pixel에서 Gemini Nano 50%+ 속도 향상 https://24-ai.news/ko/news/2026-06-27/langchain-deep-agents-prompt-caching/ 24 AI ko 2026-06-27T00:00:00Z LangChain: Deep Agents 프레임워크의 프롬프트 캐싱으로 장기 실행 에이전트 지연 시간 감소 https://24-ai.news/ko/news/2026-06-27/openai-gpt-5-6-sol-preview/ 24 AI ko 2026-06-27T00:00:00Z OpenAI: GPT-5.6 Sol 프리뷰 발표 — 코딩, 과학, 사이버보안 https://24-ai.news/hr/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI hr 2026-06-26T00:00:00Z Allen Institute: koje tokene hibridni model (OLMo 3) predviđa bolje? https://24-ai.news/hr/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI hr 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata — agentski data scientist koji stvara visokokvalitetne sintetičke podatke (Meta FAIR) https://24-ai.news/hr/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI hr 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff Tokens — pojedinačni tokeni koji okidaju neuspjeh u matematičkom reasoningu https://24-ai.news/hr/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI hr 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO — RL okvir koji nauči AI čitati emocije iz glasa, lica i teksta odjednom https://24-ai.news/hr/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI hr 2026-06-26T00:00:00Z arXiv:2606.25519: Kvantizacija napuhuje zaključivanje — skriveni trošak niskobitnih modela https://24-ai.news/hr/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI hr 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel — vanjsko izvršno usklađivanje AI agenata https://24-ai.news/hr/news/2026-06-26/claude-code-2-1-193/ 24 AI hr 2026-06-26T00:00:00Z Anthropic: Claude Code v2.1.193 — auto-mode klasifikator za shell naredbe i OpenTelemetry logiranje https://24-ai.news/hr/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI hr 2026-06-26T00:00:00Z European Commission: AWS i Microsoft Azure trebali bi biti DMA gatekeeperi https://24-ai.news/hr/news/2026-06-26/github-npm-account-protection/ 24 AI hr 2026-06-26T00:00:00Z GitHub: npm uvodi 72-satnu preventivnu zaštitu računa visokog utjecaja https://24-ai.news/hr/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI hr 2026-06-26T00:00:00Z Google Research: kako razmišljanje otključava parametarsko znanje u LLM-ovima https://24-ai.news/hr/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI hr 2026-06-26T00:00:00Z LangChain: LangSmith Fleet On-Call Copilot, Computer Use i Deep Agents RubricMiddleware https://24-ai.news/hr/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI hr 2026-06-26T00:00:00Z Microsoft: Generative Causal Testing — AI hipoteze o mozgu koje se testiraju skenerom https://24-ai.news/hr/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI hr 2026-06-26T00:00:00Z NVIDIA i AWS: EC2 G7 instance s Blackwell GPU-om donose 4,6× bolji AI inference https://24-ai.news/hr/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI hr 2026-06-26T00:00:00Z OpenAI: kako agenti mijenjaju rad — Codex 5 milijuna tjednih korisnika, rast 400% https://24-ai.news/hr/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI hr 2026-06-26T00:00:00Z PyTorch: TokenSpeed-Kernel — prijenosni high-performance kerneli za multi-silicon LLM inference https://24-ai.news/en/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI en 2026-06-26T00:00:00Z Allen Institute: which tokens does a hybrid model (OLMo 3) predict better? https://24-ai.news/en/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI en 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata — an agentic data scientist that creates high-quality synthetic data (Meta FAIR) https://24-ai.news/en/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI en 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff Tokens — single tokens that trigger failure in mathematical reasoning https://24-ai.news/en/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI en 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO — an RL framework that teaches AI to read emotions from voice, face, and text simultaneously https://24-ai.news/en/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI en 2026-06-26T00:00:00Z arXiv:2606.25519: Quantization inflates reasoning — the hidden cost of low-bit models https://24-ai.news/en/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI en 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel — external execution-time alignment for AI agents https://24-ai.news/en/news/2026-06-26/claude-code-2-1-193/ 24 AI en 2026-06-26T00:00:00Z Anthropic: Claude Code v2.1.193 — auto-mode classifier for shell commands and OpenTelemetry logging https://24-ai.news/en/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI en 2026-06-26T00:00:00Z European Commission: AWS and Microsoft Azure should be DMA gatekeepers https://24-ai.news/en/news/2026-06-26/github-npm-account-protection/ 24 AI en 2026-06-26T00:00:00Z GitHub: npm introduces 72-hour preventive protection for high-impact accounts https://24-ai.news/en/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI en 2026-06-26T00:00:00Z Google Research: how thinking unlocks parametric knowledge in LLMs https://24-ai.news/en/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI en 2026-06-26T00:00:00Z LangChain: LangSmith Fleet On-Call Copilot, Computer Use, and Deep Agents RubricMiddleware https://24-ai.news/en/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI en 2026-06-26T00:00:00Z Microsoft: Generative Causal Testing — AI hypotheses about the brain validated by scanner https://24-ai.news/en/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI en 2026-06-26T00:00:00Z NVIDIA and AWS: EC2 G7 instances with Blackwell GPU deliver 4.6× better AI inference https://24-ai.news/en/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI en 2026-06-26T00:00:00Z OpenAI: how agents are transforming work — Codex 5 million weekly users, 400% growth https://24-ai.news/en/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI en 2026-06-26T00:00:00Z PyTorch: TokenSpeed-Kernel — portable high-performance kernels for multi-silicon LLM inference https://24-ai.news/de/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI de 2026-06-26T00:00:00Z Allen Institute: Welche Token sagt ein Hybridmodell (OLMo 3) besser voraus? https://24-ai.news/de/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI de 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata — agentischer Datenwissenschaftler für hochwertige synthetische Daten (Meta FAIR) https://24-ai.news/de/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI de 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff-Token — einzelne Token, die mathematisches Schlussfolgern zum Scheitern bringen https://24-ai.news/de/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI de 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO — RL-Framework, der KI das simultane Lesen von Stimme, Gesicht und Text beibringt https://24-ai.news/de/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI de 2026-06-26T00:00:00Z arXiv:2606.25519: Quantisierung bläht das Schlussfolgern auf — verborgene Kosten von Niedrigbit-Modellen https://24-ai.news/de/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI de 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel — externes Ausführungs-Alignment für KI-Agenten https://24-ai.news/de/news/2026-06-26/claude-code-2-1-193/ 24 AI de 2026-06-26T00:00:00Z Anthropic: Claude Code v2.1.193 — Auto-Mode-Klassifikator für Shell-Befehle und OpenTelemetry-Logging https://24-ai.news/de/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI de 2026-06-26T00:00:00Z Europäische Kommission: AWS und Microsoft Azure sollen DMA-Gatekeeper werden https://24-ai.news/de/news/2026-06-26/github-npm-account-protection/ 24 AI de 2026-06-26T00:00:00Z GitHub: npm führt 72-stündigen Präventivschutz für Konten mit hoher Reichweite ein https://24-ai.news/de/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI de 2026-06-26T00:00:00Z Google Research: Wie Denken das parametrische Wissen in LLMs erschließt https://24-ai.news/de/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI de 2026-06-26T00:00:00Z LangChain: LangSmith Fleet On-Call Copilot, Computer Use und Deep Agents RubricMiddleware https://24-ai.news/de/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI de 2026-06-26T00:00:00Z Microsoft: Generative Causal Testing — KI-Hypothesen über das Gehirn, mit dem Scanner getestet https://24-ai.news/de/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI de 2026-06-26T00:00:00Z NVIDIA und AWS: EC2 G7-Instanzen mit Blackwell-GPU bringen 4,6× bessere KI-Inferenz https://24-ai.news/de/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI de 2026-06-26T00:00:00Z OpenAI: Wie Agenten die Arbeit transformieren — Codex mit 5 Millionen Wochennutzern, Wachstum 400% https://24-ai.news/de/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI de 2026-06-26T00:00:00Z PyTorch: TokenSpeed-Kernel — portable Hochleistungs-Kernel für Multi-Silicon-LLM-Inferenz https://24-ai.news/zh/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI zh 2026-06-26T00:00:00Z Allen Institute：混合模型（OLMo 3）对哪类词元预测更准？ https://24-ai.news/zh/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI zh 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata——Meta FAIR开发的智能体数据科学家，可自动生成高质量合成数据 https://24-ai.news/zh/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI zh 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff Tokens——触发数学推理失败的单个词元 https://24-ai.news/zh/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI zh 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO——让AI同时从声音、表情和文本中读懂情绪的强化学习框架 https://24-ai.news/zh/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI zh 2026-06-26T00:00:00Z arXiv:2606.25519: 量化使推理膨胀——低位模型的隐性代价 https://24-ai.news/zh/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI zh 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel——AI智能体的外部执行对齐 https://24-ai.news/zh/news/2026-06-26/claude-code-2-1-193/ 24 AI zh 2026-06-26T00:00:00Z Anthropic：Claude Code v2.1.193——Shell命令自动模式分类器与OpenTelemetry日志记录 https://24-ai.news/zh/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI zh 2026-06-26T00:00:00Z 欧洲委员会：AWS和Microsoft Azure应被列为DMA守门人 https://24-ai.news/zh/news/2026-06-26/github-npm-account-protection/ 24 AI zh 2026-06-26T00:00:00Z GitHub：npm为高影响力账户引入72小时预防性保护 https://24-ai.news/zh/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI zh 2026-06-26T00:00:00Z Google Research：推理如何解锁LLM中的参数化知识 https://24-ai.news/zh/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI zh 2026-06-26T00:00:00Z LangChain：LangSmith Fleet On-Call Copilot、Computer Use与Deep Agents RubricMiddleware https://24-ai.news/zh/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI zh 2026-06-26T00:00:00Z 微软：生成式因果测试——用AI假设通过扫描仪检验大脑 https://24-ai.news/zh/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI zh 2026-06-26T00:00:00Z 英伟达与AWS：搭载Blackwell GPU的EC2 G7实例带来4.6倍AI推理性能提升 https://24-ai.news/zh/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI zh 2026-06-26T00:00:00Z OpenAI：智能体如何改变工作——Codex每周500万用户，增长400% https://24-ai.news/zh/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI zh 2026-06-26T00:00:00Z PyTorch：TokenSpeed-Kernel——适用于多硅LLM推理的可移植高性能内核 https://24-ai.news/ja/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI ja 2026-06-26T00:00:00Z Allen Institute：ハイブリッドモデル（OLMo 3）はどのトークンをより正確に予測するか？ https://24-ai.news/ja/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI ja 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata — データサイエンティストとして高品質な合成データを生成するAIエージェント（Meta FAIR） https://24-ai.news/ja/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI ja 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff Tokens — 数学的推論の失敗を引き起こす単一トークン https://24-ai.news/ja/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI ja 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO — 声・表情・テキストから同時に感情を読み取るAIのRLフレームワーク https://24-ai.news/ja/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI ja 2026-06-26T00:00:00Z arXiv:2606.25519: 量子化が推論を膨張させる — 低ビットモデルの隠れたコスト https://24-ai.news/ja/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI ja 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel — AIエージェントの外部実行時アライメント https://24-ai.news/ja/news/2026-06-26/claude-code-2-1-193/ 24 AI ja 2026-06-26T00:00:00Z Anthropic：Claude Code v2.1.193 — シェルコマンド用オートモード分類器とOpenTelemetryロギング https://24-ai.news/ja/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI ja 2026-06-26T00:00:00Z 欧州委員会：AWSとMicrosoft AzureはDMAゲートキーパーであるべき https://24-ai.news/ja/news/2026-06-26/github-npm-account-protection/ 24 AI ja 2026-06-26T00:00:00Z GitHub：npmが高インパクトアカウントに72時間の予防的保護を導入 https://24-ai.news/ja/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI ja 2026-06-26T00:00:00Z Google Research：推論がLLMのパラメトリック知識を引き出す方法 https://24-ai.news/ja/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI ja 2026-06-26T00:00:00Z LangChain：LangSmith Fleet On-Call Copilot、Computer Use、Deep Agents RubricMiddleware https://24-ai.news/ja/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI ja 2026-06-26T00:00:00Z Microsoft：Generative Causal Testing — スキャナーでテストされるAIによる脳の仮説 https://24-ai.news/ja/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI ja 2026-06-26T00:00:00Z NVIDIAとAWS：EC2 G7インスタンスがBlackwell GPUでAI推論を4.6倍向上 https://24-ai.news/ja/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI ja 2026-06-26T00:00:00Z OpenAI：エージェントが仕事を変える方法 — Codexが週500万ユーザー、400%成長 https://24-ai.news/ja/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI ja 2026-06-26T00:00:00Z PyTorch：TokenSpeed-Kernel — マルチシリコンLLM推論のための可搬性の高いハイパフォーマンスカーネル https://24-ai.news/ko/news/2026-06-26/ai2-hybrid-token-prediction/ 24 AI ko 2026-06-26T00:00:00Z Allen Institute: 하이브리드 모델(OLMo 3)이 더 잘 예측하는 토큰은? https://24-ai.news/ko/news/2026-06-26/arxiv-autodata-agentic-data-scientist/ 24 AI ko 2026-06-26T00:00:00Z arXiv:2606.25996: Autodata — 고품질 합성 데이터를 생성하는 에이전트형 데이터 과학자(Meta FAIR) https://24-ai.news/ko/news/2026-06-26/arxiv-cliff-tokens-reasoning/ 24 AI ko 2026-06-26T00:00:00Z arXiv:2606.25524: Cliff 토큰 — 수학적 추론 실패를 유발하는 단일 토큰 https://24-ai.news/ko/news/2026-06-26/arxiv-oppo-multimodal-emotion/ 24 AI ko 2026-06-26T00:00:00Z arXiv:2606.25325: OPPO — 음성·표정·텍스트에서 동시에 감정을 읽는 RL 프레임워크 https://24-ai.news/ko/news/2026-06-26/arxiv-quantization-token-inflation/ 24 AI ko 2026-06-26T00:00:00Z arXiv:2606.25519: 양자화가 추론을 부풀린다 — 저비트 모델의 숨겨진 비용 https://24-ai.news/ko/news/2026-06-26/arxiv-unfireable-safety-kernel/ 24 AI ko 2026-06-26T00:00:00Z arXiv:2606.26057: The Unfireable Safety Kernel — AI 에이전트의 외부 실행 시점 정렬 https://24-ai.news/ko/news/2026-06-26/claude-code-2-1-193/ 24 AI ko 2026-06-26T00:00:00Z Anthropic: Claude Code v2.1.193 — 쉘 명령 자동 모드 분류기와 OpenTelemetry 로깅 https://24-ai.news/ko/news/2026-06-26/ec-dma-aws-azure-gatekeeper/ 24 AI ko 2026-06-26T00:00:00Z 유럽 집행위원회: AWS와 Microsoft Azure, DMA 게이트키퍼 지정 예비 결정 https://24-ai.news/ko/news/2026-06-26/github-npm-account-protection/ 24 AI ko 2026-06-26T00:00:00Z GitHub: npm, 고영향 계정에 72시간 예방적 계정 보호 도입 https://24-ai.news/ko/news/2026-06-26/google-thinking-to-recall-parametric-memory/ 24 AI ko 2026-06-26T00:00:00Z Google Research: 추론이 LLM의 매개변수 지식을 어떻게 활성화하는가 https://24-ai.news/ko/news/2026-06-26/langchain-june-newsletter-fleet-copilot/ 24 AI ko 2026-06-26T00:00:00Z LangChain: LangSmith Fleet On-Call Copilot, Computer Use와 Deep Agents RubricMiddleware https://24-ai.news/ko/news/2026-06-26/microsoft-gct-brain-explanations/ 24 AI ko 2026-06-26T00:00:00Z Microsoft: Generative Causal Testing — 스캐너로 검증하는 AI 뇌 가설 https://24-ai.news/ko/news/2026-06-26/nvidia-aws-ec2-g7-blackwell-cuvs/ 24 AI ko 2026-06-26T00:00:00Z NVIDIA와 AWS: EC2 G7 Blackwell GPU 인스턴스, AI 추론 4.6배 향상 https://24-ai.news/ko/news/2026-06-26/openai-agents-transforming-work-codex/ 24 AI ko 2026-06-26T00:00:00Z OpenAI: 에이전트가 일을 바꾸는 방법 — Codex 주간 사용자 500만, 400% 성장 https://24-ai.news/ko/news/2026-06-26/pytorch-tokenspeed-multi-silicon-kernels/ 24 AI ko 2026-06-26T00:00:00Z PyTorch: TokenSpeed-Kernel — 멀티 실리콘 LLM 추론을 위한 이식 가능한 고성능 커널 https://24-ai.news/hr/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI hr 2026-06-25T00:00:00Z AMD: ATOM optimizer — DP Attention i Two-Batch Overlap za DeepSeek-V4 na MI355X https://24-ai.news/hr/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI hr 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — otvoreni recepti za treniranje agentskih modela https://24-ai.news/hr/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI hr 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — specijalizirani 32B reasoning LLM ubrzava dijagnozu rijetkih bolesti u RCT-u https://24-ai.news/hr/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI hr 2026-06-25T00:00:00Z arXiv:2606.24014: RL treniranje na zdravstvenom domenu prenosi usklađenost na 80%+ OOD benchmarka https://24-ai.news/hr/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI hr 2026-06-25T00:00:00Z AWS: Huntington Bank redaktirao PII iz 400 milijuna dokumenata uz 95% točnost https://24-ai.news/hr/news/2026-06-25/claude-code-2-1-191/ 24 AI hr 2026-06-25T00:00:00Z Anthropic: Claude Code v2.1.191 — /rewind naredba, 37% manje CPU-a, MCP retry logika https://24-ai.news/hr/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI hr 2026-06-25T00:00:00Z Google DeepMind: AI Control Roadmap — defense-in-depth za sigurnost AI agenata https://24-ai.news/hr/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI hr 2026-06-25T00:00:00Z Google: DiffusionGemma 26B — 4× brža generacija teksta difuzijskim pristupom https://24-ai.news/hr/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI hr 2026-06-25T00:00:00Z Google: Computer Use u Gemini 3.5 Flash — agenti za browser, mobitel i desktop https://24-ai.news/hr/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI hr 2026-06-25T00:00:00Z Google: Gemini 3.5 Live Translate — prijevod govor-u-govor u 70+ jezika u stvarnom vremenu https://24-ai.news/hr/news/2026-06-25/github-self-service-credential-revocation/ 24 AI hr 2026-06-25T00:00:00Z GitHub: Self-service opoziv pristupnih podataka — break-glass za incident response https://24-ai.news/hr/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI hr 2026-06-25T00:00:00Z IBM: IBM, Red Hat i Palo Alto Networks šire Project Lightwell za trenutni odgovor na ranjivosti softvera https://24-ai.news/hr/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI hr 2026-06-25T00:00:00Z LangChain: kako dati AI agentu memoriju — capture, analyze, update preko LangSmitha https://24-ai.news/hr/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI hr 2026-06-25T00:00:00Z Microsoft: Talos — open-source automatska ponovna analiza genoma za rijetke bolesti https://24-ai.news/hr/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI hr 2026-06-25T00:00:00Z OpenAI: Jalapeño — vlastiti ASIC čip za LLM inference koji smanjuje ovisnost o NVIDIA-i https://24-ai.news/en/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI en 2026-06-25T00:00:00Z AMD: ATOM optimizer — DP Attention and Two-Batch Overlap for DeepSeek-V4 on MI355X https://24-ai.news/en/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI en 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — open data recipes for training agentic models https://24-ai.news/en/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI en 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — specialized 32B reasoning LLM accelerates rare disease diagnosis in RCT https://24-ai.news/en/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI en 2026-06-25T00:00:00Z arXiv:2606.24014: RL training on health domain transfers alignment to 80%+ OOD benchmarks https://24-ai.news/en/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI en 2026-06-25T00:00:00Z AWS: Huntington Bank redacted PII from 400 million documents at 95% accuracy https://24-ai.news/en/news/2026-06-25/claude-code-2-1-191/ 24 AI en 2026-06-25T00:00:00Z Anthropic: Claude Code v2.1.191 — /rewind command, 37% less CPU, MCP retry logic https://24-ai.news/en/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI en 2026-06-25T00:00:00Z Google DeepMind: AI Control Roadmap — defense-in-depth for AI agent security https://24-ai.news/en/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI en 2026-06-25T00:00:00Z Google: DiffusionGemma 26B — 4× faster text generation via diffusion approach https://24-ai.news/en/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI en 2026-06-25T00:00:00Z Google: Computer Use in Gemini 3.5 Flash — agents for browser, mobile, and desktop https://24-ai.news/en/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI en 2026-06-25T00:00:00Z Google: Gemini 3.5 Live Translate — speech-to-speech in 70+ languages in real time https://24-ai.news/en/news/2026-06-25/github-self-service-credential-revocation/ 24 AI en 2026-06-25T00:00:00Z GitHub: Self-service credential revocation — break-glass for incident response https://24-ai.news/en/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI en 2026-06-25T00:00:00Z IBM: IBM, Red Hat, and Palo Alto Networks expand Project Lightwell for instant software vulnerability response https://24-ai.news/en/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI en 2026-06-25T00:00:00Z LangChain: how to give an AI agent memory — capture, analyze, update via LangSmith https://24-ai.news/en/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI en 2026-06-25T00:00:00Z Microsoft: Talos — open-source automated iterative genomic reanalysis for rare diseases https://24-ai.news/en/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI en 2026-06-25T00:00:00Z OpenAI: Jalapeño — custom ASIC chip for LLM inference reducing NVIDIA dependence https://24-ai.news/de/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI de 2026-06-25T00:00:00Z AMD: ATOM-Optimizer — DP Attention und Two-Batch Overlap für DeepSeek-V4 auf MI355X https://24-ai.news/de/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI de 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — offene Rezepte für das Training agentischer Modelle https://24-ai.news/de/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI de 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — spezialisiertes 32B Reasoning-LLM beschleunigt Seltene-Krankheiten-Diagnose im RCT https://24-ai.news/de/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI de 2026-06-25T00:00:00Z arXiv:2606.24014: RL-Training im Gesundheitsbereich überträgt Alignment auf 80 %+ OOD-Benchmarks https://24-ai.news/de/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI de 2026-06-25T00:00:00Z AWS: Huntington Bank schwärzte PII aus 400 Millionen Dokumenten mit 95 % Genauigkeit https://24-ai.news/de/news/2026-06-25/claude-code-2-1-191/ 24 AI de 2026-06-25T00:00:00Z Anthropic: Claude Code v2.1.191 — /rewind-Befehl, 37 % weniger CPU, MCP-Retry-Logik https://24-ai.news/de/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI de 2026-06-25T00:00:00Z Google DeepMind: AI Control Roadmap — Defense-in-Depth für die Sicherheit von KI-Agenten https://24-ai.news/de/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI de 2026-06-25T00:00:00Z Google: DiffusionGemma 26B — 4× schnellere Textgenerierung durch Diffusionsansatz https://24-ai.news/de/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI de 2026-06-25T00:00:00Z Google: Computer Use in Gemini 3.5 Flash — KI-Agenten für Browser, Mobilgerät und Desktop https://24-ai.news/de/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI de 2026-06-25T00:00:00Z Google: Gemini 3.5 Live Translate — Sprach-zu-Sprach-Übersetzung in 70+ Sprachen in Echtzeit https://24-ai.news/de/news/2026-06-25/github-self-service-credential-revocation/ 24 AI de 2026-06-25T00:00:00Z GitHub: Self-Service-Widerruf von Zugangsdaten — Break-Glass für Incident Response https://24-ai.news/de/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI de 2026-06-25T00:00:00Z IBM: IBM, Red Hat und Palo Alto Networks erweitern Project Lightwell für sofortige Reaktion auf Sicherheitslücken https://24-ai.news/de/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI de 2026-06-25T00:00:00Z LangChain: Wie man einem KI-Agenten Gedächtnis gibt — Capture, Analyze, Update über LangSmith https://24-ai.news/de/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI de 2026-06-25T00:00:00Z Microsoft: Talos — Open-Source-System für automatische iterative Genomreanalyse bei seltenen Krankheiten https://24-ai.news/de/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI de 2026-06-25T00:00:00Z OpenAI: Jalapeño — eigener ASIC-Chip für LLM-Inferenz, um die Abhängigkeit von NVIDIA zu reduzieren https://24-ai.news/zh/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI zh 2026-06-25T00:00:00Z AMD: ATOM优化器 — MI355X上的DeepSeek-V4的DP注意力与双批次重叠 https://24-ai.news/zh/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI zh 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — 智能体模型训练的开放数据方案 https://24-ai.news/zh/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI zh 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — 专用32B推理LLM在RCT中加速罕见病诊断 https://24-ai.news/zh/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI zh 2026-06-25T00:00:00Z arXiv:2606.24014: 健康领域RL训练将对齐能力迁移至80%以上的OOD基准 https://24-ai.news/zh/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI zh 2026-06-25T00:00:00Z AWS：Huntington Bank从4亿份文档中清除PII，准确率达95% https://24-ai.news/zh/news/2026-06-25/claude-code-2-1-191/ 24 AI zh 2026-06-25T00:00:00Z Anthropic：Claude Code v2.1.191 — /rewind命令、CPU减少37%、MCP重试逻辑 https://24-ai.news/zh/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI zh 2026-06-25T00:00:00Z Google DeepMind：AI控制路线图 — 智能体安全的纵深防御 https://24-ai.news/zh/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI zh 2026-06-25T00:00:00Z Google：DiffusionGemma 26B — 扩散方法实现4倍更快的文本生成 https://24-ai.news/zh/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI zh 2026-06-25T00:00:00Z Google：Gemini 3.5 Flash中的Computer Use — 浏览器、手机与桌面智能体 https://24-ai.news/zh/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI zh 2026-06-25T00:00:00Z Google：Gemini 3.5 Live Translate — 实时70+语言语音互译 https://24-ai.news/zh/news/2026-06-25/github-self-service-credential-revocation/ 24 AI zh 2026-06-25T00:00:00Z GitHub：自助凭据撤销 — 事件响应的紧急访问控制 https://24-ai.news/zh/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI zh 2026-06-25T00:00:00Z IBM：IBM、Red Hat与Palo Alto Networks扩展Project Lightwell以即时响应软件漏洞 https://24-ai.news/zh/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI zh 2026-06-25T00:00:00Z LangChain：如何为AI智能体赋予记忆 — 通过LangSmith实现捕获、分析与更新 https://24-ai.news/zh/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI zh 2026-06-25T00:00:00Z Microsoft：Talos — 罕见病基因组自动迭代重分析开源系统 https://24-ai.news/zh/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI zh 2026-06-25T00:00:00Z OpenAI：Jalapeño — 减少对NVIDIA依赖的自研LLM推理ASIC芯片 https://24-ai.news/ja/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI ja 2026-06-25T00:00:00Z AMD: ATOMオプティマイザー — MI355X上でDeepSeek-V4向けDP AttentionとTwo-Batch Overlap https://24-ai.news/ja/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI ja 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — エージェントモデル訓練のためのオープンデータレシピ https://24-ai.news/ja/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI ja 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — 特化型32B推論LLMがRCTで希少疾患診断を加速 https://24-ai.news/ja/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI ja 2026-06-25T00:00:00Z arXiv:2606.24014: 医療ドメインのRL訓練が50以上のOODベンチマークの80%以上でアライメントを転移 https://24-ai.news/ja/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI ja 2026-06-25T00:00:00Z AWS: Huntington Bankが4億件以上の文書からPIIを95%の精度で削除 https://24-ai.news/ja/news/2026-06-25/claude-code-2-1-191/ 24 AI ja 2026-06-25T00:00:00Z Anthropic: Claude Code v2.1.191 — /rewindコマンド、CPU使用率37%削減、MCPリトライロジック https://24-ai.news/ja/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI ja 2026-06-25T00:00:00Z Google DeepMind: AI Control Roadmap — AIエージェントセキュリティの多層防御 https://24-ai.news/ja/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI ja 2026-06-25T00:00:00Z Google: DiffusionGemma 26B — 拡散アプローチによる4倍高速なテキスト生成 https://24-ai.news/ja/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI ja 2026-06-25T00:00:00Z Google: Gemini 3.5 FlashにComputer Use — ブラウザ・スマホ・デスクトップ対応エージェント https://24-ai.news/ja/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI ja 2026-06-25T00:00:00Z Google: Gemini 3.5 Live Translate — 70言語以上でリアルタイム音声翻訳 https://24-ai.news/ja/news/2026-06-25/github-self-service-credential-revocation/ 24 AI ja 2026-06-25T00:00:00Z GitHub: セルフサービス認証情報失効 — インシデント対応のためのbreak-glass機能 https://24-ai.news/ja/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI ja 2026-06-25T00:00:00Z IBM: IBM・Red Hat・Palo Alto Networksがソフトウェア脆弱性への迅速対応でProject Lightwellを拡大 https://24-ai.news/ja/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI ja 2026-06-25T00:00:00Z LangChain: AIエージェントに記憶を持たせる方法 — LangSmithでcapture・analyze・update https://24-ai.news/ja/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI ja 2026-06-25T00:00:00Z Microsoft: Talos — 希少疾患向けオープンソース自動ゲノム再解析システム https://24-ai.news/ja/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI ja 2026-06-25T00:00:00Z OpenAI: JalapeñoチップをBroadcomと共同開発 — NVIDIAへの依存から脱却を図るカスタムASIC https://24-ai.news/ko/news/2026-06-25/amd-atom-deepseek-v4-mi355x/ 24 AI ko 2026-06-25T00:00:00Z AMD: ATOM 옵티마이저 — MI355X에서 DeepSeek-V4를 위한 DP Attention 및 Two-Batch Overlap https://24-ai.news/ko/news/2026-06-25/arxiv-openthoughts-agent-data-recipes/ 24 AI ko 2026-06-25T00:00:00Z arXiv:2606.24855: OpenThoughts-Agent — 에이전트 모델 훈련을 위한 오픈 데이터 레시피 https://24-ai.news/ko/news/2026-06-25/arxiv-radar-rare-disease-reasoning-llm/ 24 AI ko 2026-06-25T00:00:00Z arXiv:2606.24510: RaDaR — 특화된 32B 추론 LLM, RCT에서 희귀 질환 진단 가속화 https://24-ai.news/ko/news/2026-06-25/arxiv-rl-broadly-beneficial-alignment-transfer/ 24 AI ko 2026-06-25T00:00:00Z arXiv:2606.24014: 의료 도메인 RL 훈련, 50개 이상 OOD 벤치마크의 80% 이상에서 정렬 향상 https://24-ai.news/ko/news/2026-06-25/aws-huntington-bank-pii-redaction/ 24 AI ko 2026-06-25T00:00:00Z AWS: Huntington Bank, 4억 건 문서에서 PII를 95% 정확도로 삭제 https://24-ai.news/ko/news/2026-06-25/claude-code-2-1-191/ 24 AI ko 2026-06-25T00:00:00Z Anthropic: Claude Code v2.1.191 — /rewind 명령어, CPU 37% 절감, MCP 재시도 로직 https://24-ai.news/ko/news/2026-06-25/deepmind-ai-control-roadmap-agent-security/ 24 AI ko 2026-06-25T00:00:00Z Google DeepMind: AI Control Roadmap — 에이전트 보안을 위한 심층 방어 https://24-ai.news/ko/news/2026-06-25/diffusion-gemma-26b-faster-text-generation/ 24 AI ko 2026-06-25T00:00:00Z Google: DiffusionGemma 26B — 확산 방식으로 4배 빠른 텍스트 생성 https://24-ai.news/ko/news/2026-06-25/gemini-3-5-flash-computer-use/ 24 AI ko 2026-06-25T00:00:00Z Google: Gemini 3.5 Flash에 Computer Use — 브라우저, 모바일, 데스크톱 에이전트 https://24-ai.news/ko/news/2026-06-25/gemini-live-3-5-translate-70-languages/ 24 AI ko 2026-06-25T00:00:00Z Google: Gemini 3.5 Live Translate — 실시간 70개 이상 언어 음성 통역 https://24-ai.news/ko/news/2026-06-25/github-self-service-credential-revocation/ 24 AI ko 2026-06-25T00:00:00Z GitHub: 셀프서비스 자격 증명 해지 — 인시던트 대응을 위한 break-glass https://24-ai.news/ko/news/2026-06-25/ibm-redhat-palo-alto-project-lightwell/ 24 AI ko 2026-06-25T00:00:00Z IBM: IBM, Red Hat, Palo Alto Networks, 소프트웨어 취약점 즉각 대응을 위한 Project Lightwell 확장 https://24-ai.news/ko/news/2026-06-25/langchain-agent-memory-langsmith/ 24 AI ko 2026-06-25T00:00:00Z LangChain: LangSmith를 통한 AI 에이전트 메모리 구현 — capture, analyze, update https://24-ai.news/ko/news/2026-06-25/microsoft-talos-genomic-reanalysis-rare-disease/ 24 AI ko 2026-06-25T00:00:00Z Microsoft: Talos — 희귀 질환을 위한 오픈소스 자동 반복 유전체 재분석 https://24-ai.news/ko/news/2026-06-25/openai-broadcom-jalapeno-inference-chip/ 24 AI ko 2026-06-25T00:00:00Z OpenAI: Jalapeño — NVIDIA 의존도를 줄이는 LLM 추론용 자체 ASIC 칩 https://24-ai.news/hr/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI hr 2026-06-24T00:00:00Z AMD ROCm: EAGLE3 speculative decoding ubrzava Kimi-K2.5 za 33% na MI325X https://24-ai.news/hr/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI hr 2026-06-24T00:00:00Z Anthropic: Claude Code v2.1.187 — sandbox credentials zaštita, org restrikcije modela, CJK fix https://24-ai.news/hr/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI hr 2026-06-24T00:00:00Z Anthropic: Claude Tag donosi multiplayer AI agenta direktno u Slack timove https://24-ai.news/hr/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI hr 2026-06-24T00:00:00Z arXiv:2606.23189: 11 od 15 AI agenata curi privatne podatke u više od pola scenarija https://24-ai.news/hr/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI hr 2026-06-24T00:00:00Z arXiv:2606.23181: DART — bez treninga do adaptivnog mišljenja u hibridnim reasoning modelima https://24-ai.news/hr/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI hr 2026-06-24T00:00:00Z arXiv:2606.22844: RaMem rješava kolaps konteksta u dugoročnoj agentskoj memoriji https://24-ai.news/hr/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI hr 2026-06-24T00:00:00Z AWS: Bedrock AgentCore pool-model multi-tenancy — dijeljena infra, izolirani tenanti https://24-ai.news/hr/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI hr 2026-06-24T00:00:00Z CNCF: Agent Auth — siguran identitet i delegacija ovlasti za AI agente https://24-ai.news/hr/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI hr 2026-06-24T00:00:00Z GitHub: koalicija traži izmjene California AI Transparency Acta radi zaštite open-sourcea https://24-ai.news/hr/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI hr 2026-06-24T00:00:00Z GitHub: Copilot dobiva novo CLI terminal sučelje i podršku za vlastiti API ključ (BYOK) https://24-ai.news/hr/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI hr 2026-06-24T00:00:00Z Mistral: OCR 4 — strukturirana ekstrakcija dokumenata s bounding boxovima u 170 jezika https://24-ai.news/hr/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI hr 2026-06-24T00:00:00Z NVIDIA: Agent Toolkit — otvorena osnova za enterprise AI agente s Nemotron modelima https://24-ai.news/hr/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI hr 2026-06-24T00:00:00Z OpenAI: Pridruživanje Appia Foundationu za zajedničke standarde evaluacije naprednog AI-ja https://24-ai.news/hr/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI hr 2026-06-24T00:00:00Z OpenAI: GPT-5 Pro pomogao riješiti 3-godišnju imunološku misteriju o T-stanicama https://24-ai.news/hr/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI hr 2026-06-24T00:00:00Z PyTorch/SGLang: DeepSeek-V4 Pro na NVIDIA GB300 — 5× veći throughput uz isti interaktivitet https://24-ai.news/en/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI en 2026-06-24T00:00:00Z AMD ROCm: EAGLE3 speculative decoding speeds up Kimi-K2.5 by 33% on MI325X https://24-ai.news/en/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI en 2026-06-24T00:00:00Z Anthropic: Claude Code v2.1.187 — sandbox credentials protection, org model restrictions, CJK fix https://24-ai.news/en/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI en 2026-06-24T00:00:00Z Anthropic: Claude Tag brings a multiplayer AI agent directly into Slack teams https://24-ai.news/en/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI en 2026-06-24T00:00:00Z arXiv:2606.23189: 11 of 15 AI agents leak private data in more than half of scenarios https://24-ai.news/en/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI en 2026-06-24T00:00:00Z arXiv:2606.23181: DART — training-free adaptive thinking in hybrid reasoning models https://24-ai.news/en/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI en 2026-06-24T00:00:00Z arXiv:2606.22844: RaMem solves context collapse in long-term agentic memory https://24-ai.news/en/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI en 2026-06-24T00:00:00Z AWS: Bedrock AgentCore pool-model multi-tenancy — shared infrastructure, isolated tenants https://24-ai.news/en/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI en 2026-06-24T00:00:00Z CNCF: Agent Auth — secure identity and authority delegation for AI agents https://24-ai.news/en/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI en 2026-06-24T00:00:00Z GitHub: coalition demands changes to California AI Transparency Act to protect open source https://24-ai.news/en/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI en 2026-06-24T00:00:00Z GitHub: Copilot gets a new CLI terminal interface and BYOK support https://24-ai.news/en/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI en 2026-06-24T00:00:00Z Mistral: OCR 4 — structured document extraction with bounding boxes in 170 languages https://24-ai.news/en/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI en 2026-06-24T00:00:00Z NVIDIA: Agent Toolkit — open foundation for enterprise AI agents with Nemotron models https://24-ai.news/en/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI en 2026-06-24T00:00:00Z OpenAI: Joining Appia Foundation to build shared standards for advanced AI https://24-ai.news/en/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI en 2026-06-24T00:00:00Z OpenAI: GPT-5 Pro helped solve a 3-year immunology mystery about T-cells https://24-ai.news/en/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI en 2026-06-24T00:00:00Z PyTorch/SGLang: DeepSeek-V4 Pro on NVIDIA GB300 — 5× higher throughput with the same interactivity https://24-ai.news/de/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI de 2026-06-24T00:00:00Z AMD ROCm: EAGLE3 Speculative Decoding beschleunigt Kimi-K2.5 um 33 % auf MI325X https://24-ai.news/de/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI de 2026-06-24T00:00:00Z Anthropic: Claude Code v2.1.187 — Sandbox-Zugangsdatenschutz, Organisationsmodell-Einschränkungen, CJK-Fix https://24-ai.news/de/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI de 2026-06-24T00:00:00Z Anthropic: Claude Tag bringt einen Multiplayer-KI-Agenten direkt in Slack-Teams https://24-ai.news/de/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI de 2026-06-24T00:00:00Z arXiv:2606.23189: 11 von 15 KI-Agenten geben private Daten in mehr als der Hälfte der Szenarien weiter https://24-ai.news/de/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI de 2026-06-24T00:00:00Z arXiv:2606.23181: DART — adaptives Denken in hybriden Reasoning-Modellen ohne Training https://24-ai.news/de/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI de 2026-06-24T00:00:00Z arXiv:2606.22844: RaMem löst Kontextkollaps im Langzeitgedächtnis agentischer Systeme https://24-ai.news/de/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI de 2026-06-24T00:00:00Z AWS: Bedrock AgentCore Pool-Modell-Mandantenfähigkeit — gemeinsame Infrastruktur, isolierte Mandanten https://24-ai.news/de/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI de 2026-06-24T00:00:00Z CNCF: Agent Auth — sichere Identität und Befugnisdelegation für KI-Agenten https://24-ai.news/de/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI de 2026-06-24T00:00:00Z GitHub: Koalition fordert Änderungen am California AI Transparency Act zum Schutz von Open Source https://24-ai.news/de/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI de 2026-06-24T00:00:00Z GitHub: Copilot erhält neue CLI-Terminal-Oberfläche und Unterstützung für eigene API-Schlüssel (BYOK) https://24-ai.news/de/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI de 2026-06-24T00:00:00Z Mistral: OCR 4 — strukturierte Dokumentenextraktion mit Bounding Boxes in 170 Sprachen https://24-ai.news/de/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI de 2026-06-24T00:00:00Z NVIDIA: Agent Toolkit — offene Grundlage für Enterprise-KI-Agenten mit Nemotron-Modellen https://24-ai.news/de/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI de 2026-06-24T00:00:00Z OpenAI: Beitritt zur Appia Foundation für gemeinsame Standards zur Bewertung fortgeschrittener KI https://24-ai.news/de/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI de 2026-06-24T00:00:00Z OpenAI: GPT-5 Pro half, ein 3-jähriges immunologisches Rätsel über T-Zellen zu lösen https://24-ai.news/de/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI de 2026-06-24T00:00:00Z PyTorch/SGLang: DeepSeek-V4 Pro auf NVIDIA GB300 — 5× höherer Durchsatz bei gleichem Interaktivitätsniveau https://24-ai.news/zh/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI zh 2026-06-24T00:00:00Z AMD ROCm：EAGLE3推测解码在MI325X上将Kimi-K2.5加速33% https://24-ai.news/zh/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI zh 2026-06-24T00:00:00Z Anthropic：Claude Code v2.1.187——沙盒凭证保护、组织模型限制、CJK修复 https://24-ai.news/zh/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI zh 2026-06-24T00:00:00Z Anthropic：Claude Tag将多人AI助手直接引入Slack团队 https://24-ai.news/zh/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI zh 2026-06-24T00:00:00Z arXiv:2606.23189：15个AI Agent中有11个在超过半数场景下泄露私人数据 https://24-ai.news/zh/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI zh 2026-06-24T00:00:00Z arXiv:2606.23181：DART——无需训练的混合推理模型自适应思考预算 https://24-ai.news/zh/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI zh 2026-06-24T00:00:00Z arXiv:2606.22844：RaMem解决长期Agent记忆中的上下文坍塌问题 https://24-ai.news/zh/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI zh 2026-06-24T00:00:00Z AWS：Bedrock AgentCore池模型多租户——共享基础设施，隔离租户 https://24-ai.news/zh/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI zh 2026-06-24T00:00:00Z CNCF：Agent Auth——AI Agent的安全身份与权限委托 https://24-ai.news/zh/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI zh 2026-06-24T00:00:00Z GitHub：联盟要求修改加利福尼亚AI透明度法案以保护开源 https://24-ai.news/zh/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI zh 2026-06-24T00:00:00Z GitHub：Copilot获得新CLI终端界面和自带API密钥支持（BYOK） https://24-ai.news/zh/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI zh 2026-06-24T00:00:00Z Mistral：OCR 4——支持170种语言的结构化文档提取与边界框 https://24-ai.news/zh/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI zh 2026-06-24T00:00:00Z NVIDIA：Agent Toolkit——基于Nemotron模型的企业AI Agent开放基础平台 https://24-ai.news/zh/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI zh 2026-06-24T00:00:00Z OpenAI：加入Appia Foundation，共建先进AI评估标准 https://24-ai.news/zh/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI zh 2026-06-24T00:00:00Z OpenAI：GPT-5 Pro协助解开长达3年的T细胞免疫学谜题 https://24-ai.news/zh/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI zh 2026-06-24T00:00:00Z PyTorch/SGLang：DeepSeek-V4 Pro在NVIDIA GB300上——相同交互性下吞吐量提升5倍 https://24-ai.news/ja/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI ja 2026-06-24T00:00:00Z AMD ROCm: EAGLE3投機的デコードがMI325X上でKimi-K2.5を33%高速化 https://24-ai.news/ja/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI ja 2026-06-24T00:00:00Z Anthropic: Claude Code v2.1.187 — サンドボックス認証情報保護、組織モデル制限、CJKバグ修正 https://24-ai.news/ja/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI ja 2026-06-24T00:00:00Z Anthropic: Claude TagがSlackチームに直接マルチプレイヤーAIエージェントをもたらす https://24-ai.news/ja/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI ja 2026-06-24T00:00:00Z arXiv:2606.23189: 15のAIエージェントのうち11がシナリオの半数以上でプライベートデータを漏洩 https://24-ai.news/ja/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI ja 2026-06-24T00:00:00Z arXiv:2606.23181: DART — トレーニング不要でハイブリッド推論モデルに適応的思考を実現 https://24-ai.news/ja/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI ja 2026-06-24T00:00:00Z arXiv:2606.22844: RaMem——長期エージェントメモリのコンテキスト崩壊を解決 https://24-ai.news/ja/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI ja 2026-06-24T00:00:00Z AWS: Bedrock AgentCoreプールモデルマルチテナンシー——共有インフラ、分離されたテナント https://24-ai.news/ja/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI ja 2026-06-24T00:00:00Z CNCF: Agent Auth——AIエージェントのための安全なIDと権限委任 https://24-ai.news/ja/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI ja 2026-06-24T00:00:00Z GitHub: 連合がオープンソース保護のためにカリフォルニアAI透明性法の改正を求める https://24-ai.news/ja/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI ja 2026-06-24T00:00:00Z GitHub: CopilotがBYOKサポートで新しいCLIターミナルインターフェースを一般提供 https://24-ai.news/ja/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI ja 2026-06-24T00:00:00Z Mistral: OCR 4 — 170言語でバウンディングボックスによる構造化ドキュメント抽出 https://24-ai.news/ja/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI ja 2026-06-24T00:00:00Z NVIDIA: Agent Toolkit——NemotronモデルによるエンタープライズAIエージェントのオープン基盤 https://24-ai.news/ja/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI ja 2026-06-24T00:00:00Z OpenAI: 先進AIの共通評価標準のためAppia Foundationに参加 https://24-ai.news/ja/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI ja 2026-06-24T00:00:00Z OpenAI: GPT-5 Proが3年間の免疫学的謎——T細胞の謎——を解決するのを助けた https://24-ai.news/ja/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI ja 2026-06-24T00:00:00Z PyTorch/SGLang: NVIDIA GB300上のDeepSeek-V4 Pro——同じインタラクティビティで5倍のスループット https://24-ai.news/ko/news/2026-06-24/amd-rocm-eagle3-kimi-mi325x/ 24 AI ko 2026-06-24T00:00:00Z AMD ROCm: EAGLE3 추측 디코딩으로 MI325X에서 Kimi-K2.5 33% 가속 https://24-ai.news/ko/news/2026-06-24/anthropic-claude-code-2-1-187/ 24 AI ko 2026-06-24T00:00:00Z Anthropic: Claude Code v2.1.187 — 샌드박스 자격증명 보호, 조직 모델 제한, CJK 수정 https://24-ai.news/ko/news/2026-06-24/anthropic-claude-tag-slack-team-agent/ 24 AI ko 2026-06-24T00:00:00Z Anthropic: Claude Tag로 Slack 팀에 멀티플레이어 AI 에이전트 직접 도입 https://24-ai.news/ko/news/2026-06-24/arxiv-computer-use-agents-privacy/ 24 AI ko 2026-06-24T00:00:00Z arXiv:2606.23189: AI 에이전트 15개 중 11개가 절반 이상의 시나리오에서 개인 데이터 유출 https://24-ai.news/ko/news/2026-06-24/arxiv-dart-adaptive-thinking-budget/ 24 AI ko 2026-06-24T00:00:00Z arXiv:2606.23181: DART — 훈련 없이 하이브리드 추론 모델에서 적응형 사고 구현 https://24-ai.news/ko/news/2026-06-24/arxiv-ramem-agentic-memory/ 24 AI ko 2026-06-24T00:00:00Z arXiv:2606.22844: RaMem, 장기 에이전트 메모리의 맥락 붕괴 해결 https://24-ai.news/ko/news/2026-06-24/aws-agentcore-pool-multitenancy/ 24 AI ko 2026-06-24T00:00:00Z AWS: Bedrock AgentCore 풀 모델 멀티테넌시 — 공유 인프라, 격리된 테넌트 https://24-ai.news/ko/news/2026-06-24/cncf-agent-auth-identity-delegation/ 24 AI ko 2026-06-24T00:00:00Z CNCF: Agent Auth — AI 에이전트를 위한 안전한 신원과 권한 위임 https://24-ai.news/ko/news/2026-06-24/github-california-ai-transparency-act-coalition/ 24 AI ko 2026-06-24T00:00:00Z GitHub: 연합, 오픈소스 보호를 위한 California AI Transparency Act 수정 요구 https://24-ai.news/ko/news/2026-06-24/github-copilot-cli-terminal-byok/ 24 AI ko 2026-06-24T00:00:00Z GitHub: Copilot, 새로운 CLI 터미널 인터페이스와 자체 API 키(BYOK) 지원 획득 https://24-ai.news/ko/news/2026-06-24/mistral-ocr-4-170-languages/ 24 AI ko 2026-06-24T00:00:00Z Mistral: OCR 4 — 170개 언어에서 바운딩 박스를 사용한 구조화된 문서 추출 https://24-ai.news/ko/news/2026-06-24/nvidia-agent-toolkit-nemotron-enterprise/ 24 AI ko 2026-06-24T00:00:00Z NVIDIA: Agent Toolkit — Nemotron 모델을 갖춘 엔터프라이즈 AI 에이전트를 위한 오픈 기반 https://24-ai.news/ko/news/2026-06-24/openai-appia-foundation-ai-standards/ 24 AI ko 2026-06-24T00:00:00Z OpenAI: 고급 AI를 위한 공동 표준 구축을 위해 Appia Foundation 합류 https://24-ai.news/ko/news/2026-06-24/openai-gpt5-pro-immunology-tcell/ 24 AI ko 2026-06-24T00:00:00Z OpenAI: GPT-5 Pro가 T세포에 관한 3년 면역학 미스터리 해결 지원 https://24-ai.news/ko/news/2026-06-24/pytorch-deepseek-v4-sglang-gb300/ 24 AI ko 2026-06-24T00:00:00Z PyTorch/SGLang: GB300에서 DeepSeek-V4 Pro — 동일한 인터랙티브성으로 5배 높은 처리량