Machine Learning13 min reading time
Last Week in AI #345 - 5 new models, 9 misalignment incidents, some Dots
Last Week in AI
Read the full articleAnthropic and OpenAI released multiple new AI models with improved safety features, reduced costs, and enhanced performance. Anthropic's Claude Opus 5.5 and Sonnet 5.5 focus on cybersecurity and faster processing, while OpenAI's GPT-6 Sol and Luna offer lower prices and fewer errors, with GPT-6.1 Sol nearing top-tier performance at reduced cost.
- GPT-6 Astra vs GPT-6.1 Sol vs Gemini 4 Argon vs Claude Fable 5.1: Which Frontier Model Fits Which Job· MarkTechPost
- New OpenAI Product Fails Spectacularly During Live On-Stage Demo· Futurism
- Astra 6.1 Pulled As Insufficiently Aligned· Don't Worry About the Vase
- Speaker Mike Johnson tells CNBC he hopes AI guardrails are voluntary· 2 sources
- DevDay 2026 Recap· Open AI News
- OpenAI Scraps Debut of AI Model as It Sets New Guardrails· 21 sources
