Aivio News — Latest AI News, Models & ResearchFollow global AI news, model launches, research, benchmarks, policy and industry moves from OpenAI, Anthropic, Google, xAI and more.https://aivio.news/en© 2026 GrowQ ABCrowdStrike launches SafeMind, pitting an attack model against a defence modelhttps://aivio.news/news/crowdstrike-safemind-red-tempest-blue-solano-nvidia/https://aivio.news/news/crowdstrike-safemind-red-tempest-blue-solano-nvidia/The product is not either model but the loop between them. Every performance figure in the launch is CrowdStrike's own, measured against competitors it has not named.Thu, 03 Sep 2026 09:47:00 GMTsafetysecurityai-agentsmodel-evaluationsenterprise-aiLove LindbergOpenAI says Astra is its first model to reach Critical cyber capabilityhttps://aivio.news/news/openai-astra-critical-cybersecurity-preparedness-framework/https://aivio.news/news/openai-astra-critical-cybersecurity-preparedness-framework/The model scored perfectly on ExploitBench and chained its own zero-days into a browser sandbox escape. Its full cyber capabilities will not be available at launch.Thu, 03 Sep 2026 06:20:00 GMTsafetysecuritymodel-evaluationsalignmentai-agentsLove LindbergAlibaba's Qwen3.8-Max-0902 snapshot claims the top CodeArena spothttps://aivio.news/news/alibaba-qwen3-8-max-0902-snapshot-codearena/https://aivio.news/news/alibaba-qwen3-8-max-0902-snapshot-codearena/The company shipped a dated snapshot rather than a new model — a release pattern that is quietly becoming the norm at the frontier.Wed, 02 Sep 2026 12:40:00 GMTmodelslarge-language-modelscoding-assistantsmodel-evaluationsLove LindbergAmazon triples its Nvidia order to 2 million GPUs for 2027 and 2028https://aivio.news/news/amazon-triples-nvidia-gpu-order-aws-2027-2028/https://aivio.news/news/amazon-triples-nvidia-gpu-order-aws-2027-2028/The two-year delivery horizon says more about the supply queue than the headline number does — and Amazon is scaling its own accelerators at the same time.Wed, 02 Sep 2026 09:30:00 GMThardwaregpusdata-centreschipsenergyenterprise-aiLove LindbergAnthropic ships Claude Fable 5.1 and cuts cache reads by 75%https://aivio.news/news/anthropic-claude-fable-5-1-mythos-5-1-cache-pricing/https://aivio.news/news/anthropic-claude-fable-5-1-mythos-5-1-cache-pricing/The two models are the same system under different safeguard regimes. The pricing change, not the benchmark table, is what will show up in developers' bills.Wed, 02 Sep 2026 07:15:00 GMTmodelslarge-language-modelscoding-assistantsai-agentsmodel-evaluationsLove Lindberg