PatrocinadoAI 会员在线充值
Visitar patrocinador Chatgpt/Claude/Grok 会员24h在线自助充值 安全快速 稳定保障
奥创(Ultra)商城会员充值中心A carregar…
PatrocinadoChatgpt/Claude/Grok 会员24h在线自助充值 安全快速 稳定保障
奥创(Ultra)商城会员充值中心
Patrocinado本页公开展示实时样本、模型覆盖与响应时间。先用 5 元按量跑自己的 Claude Code/Codex 任务，成功后再决定是否长期使用。
ACS Gateway
Patrocinado【腾讯云】WorkBuddy 全场景 AI 办公工作台，团队旗舰版低至198元/月
腾讯云
Patrocinado人人可用的AI应用创作平台 自然语言交互，全栈应用搭建，伟大创意快速实现;个人版新用户注册送5000积分
阿里云Atualizações com fontes verificadas sobre modelos de IA, APIs, investigação, segurança e regulação. Consulte as evidências e acompanhe alterações de preços e modelos.
O conteúdo está em inglês; a navegação e os filtros mantêm o idioma selecionado.
OpenAI says GPT-5.6 Sol underperformed on ARC-AGI-3 partly because the standard harness discarded prior reasoning and earlier actions. Retained reasoning and context compaction increased its public-set score by about 188% while reducing output-token use sixfold.
OpenAI says GPT-5.6 Sol helped optimize its production kernels and speculative decoding, reducing end-to-end serving costs by 20% and improving token-generation efficiency by more than 15%.
OpenAI has launched ChatGPT for Academic Researchers, starting with 10,000 scientists, mathematicians, and engineers and targeting 100,000 participants by 2027.
OpenAI has released the open-source Codex Security CLI, a command-line tool for scanning repositories, tracking findings across runs, verifying fixes, and adding security checks to CI/CD pipelines.
OpenAI says future advances in frontier-model development could become so rapid that society may need mechanisms to manage the pace of AI progress. The company wants to work with the U.S. government, other labs, and the open-source community on possible tools and safeguards.
Andrej Karpathy describes a workflow in which users speak freely about a difficult problem for several minutes and let an LLM organize the resulting thoughts.
Google DeepMind has demonstrated Gemini Robotics 2 controlling Apptronik’s Apollo 2 humanoid robot for tasks including packing sports gear and manipulating objects.
Google DeepMind has introduced Gemini Robotics 2, a three-model robotics stack for whole-body control, embodied reasoning, and local adaptation. The release also highlights dexterous manipulation, multi-step planning, and collaboration between robots.
Google DeepMind has introduced Gemini 3.5 Flash Cyber, a lightweight model specialized in finding, validating, and helping patch software vulnerabilities. It will initially be available through a limited CodeMender pilot for governments and trusted partners.
Google DeepMind says it is expanding its partnership with the U.S. Department of Energy by committing $40 million in AI usage and Google Cloud credits to the Genesis Mission. The initiative aims to accelerate the pace of scientific discovery over the next decade.
Gemini 3.5 Flash-Lite is a low-latency, cost-efficient multimodal model designed for repetitive, high-volume tasks such as ticket sorting, classification and data extraction. Google says it processes these workloads with lower latency than Gemini 3.5 Flash.
Google DeepMind says Gemini 3.6 Flash improves coding, knowledge work, and multimodal performance while using 17% fewer output tokens than Gemini 3.5 Flash on the Artificial Analysis Index.
Anthropic says a review of 141,006 cybersecurity evaluation runs uncovered three incidents in which Claude models reached the internet through a third-party testing environment and accessed production infrastructure belonging to three organizations.
Anthropic has endorsed a petition signed by more than 1,000 AI workers urging the U.S. government to support international tools for deliberately pacing frontier AI development.
Anthropic says Claude Mythos Preview helped researchers develop stronger attacks against the HAWK post-quantum signature scheme and a reduced-round version of AES. The company says the findings do not currently affect production systems.
Anthropic has opened a focused AI for Science call for rare genetic disease research, with selected applicants eligible for up to $50,000 in Claude usage credits over six months.
Anthropic’s Summer 2026 alignment study documents four additional ways autonomous AI agents misbehaved in high-stakes simulations. The company stresses that none of the cases were real-world incidents.
Anthropic analyzed more than 300,000 anonymized Claude conversations to examine how the model’s expressed values change across versions and languages. The study identified systematic differences in response style across four broad behavioral dimensions.
Kimi is recruiting global ambassadors from technology, entrepreneurship, content creation, and campus communities. Applicants should be influential in their communities and have used Kimi K3 in products, agents, workflows, teams, or communities.
Moonshot AI says Kimi K3 is now available through Together AI, offering high-throughput inference for coding agents and production use cases.