Claude Opus 4.8: Anthropic's Honest, Powerful AI Upgrade

Anthropic has released Claude Opus 4.8, an upgraded AI model offering stronger performance in coding and professional tasks. This new version is significantly more honest, flagging uncertainties and less likely to make unsupported claims. Opus 4.8 achieves state-of-the-art results, scoring 1890 on GDPval-AA for knowledge work and 69.2% on SWE-Bench Pro. Anthropic also introduced dynamic workflows for large tasks, where Claude breaks down complex prompts into subtasks and uses parallel sub-agents. Meanwhile, OpenAI launched Rosalind Biodefense, giving trusted developers access to GPT-Rosalind for defensive biology work like epidemiological modeling. Mistral introduced Search Toolkit, an open-source framework to streamline search pipelines for AI applications. Mistral also launched Vibe, its new live agent product and main AI interface, replacing LeChat, which includes a Work Mode and a Code Mode. These advancements show a continued push for more capable and reliable AI tools across various fields.