GPT-5
26 mentions across all digests
GPT-5 is a large language model developed by OpenAI used as a benchmark reference point; open-weight models like DeepSeek V3.2 and OpenAI's own gpt-oss-120b are evaluated against it on standard benchmarks.
Introducing GPT-5 for developers
Introducing GPT-5
GPT-5 System Card
Kevin Weil and Bill Peebles exit OpenAI as company continues to shed ‘side quests’
OpenAI shuts down Sora (burning $1M daily) and research programs as executives Kevin Weil and Bill Peebles exit, pivoting away from exploratory R&D toward profitable enterprise AI consolidation.
Our evaluation of Claude Mythos Preview’s cyber capabilities
Claude Mythos Preview became the first AI model to complete a 32-step network attack scenario (73% on expert CTF challenges), but the AI Security Institute emphasizes that controlled test environments overstate real-world cyber effectiveness.