Anthropic Splits One Claude in Two

Basically. Fable 5.1 and Mythos 5.1 use the same underlying intelligence. Fable is available to everyone, while Mythos is reserved for approved cybersecurity and life-science work.

It gets fewer restrictions on dangerous expert work. Mythos designed protein binders with a nearly 50% hit rate and optimized biology models to run up to 2.5 times faster, but only approved researchers can use it.

For starters, it makes incredible games. It also got cheaper and less trigger-happy. Anthropic says typical workloads should cost about 25% less, savings can reach 45% on heavily agentic work and its cyber safeguards now produce 60% fewer false alarms.

For younger students, yes, but only for one year. NYC’s moratorium blocks student-facing generative AI from 2-K through eighth grade and covers nearly 600,000 students.

Teachers can use approved AI for planning. High schools get critical-thinking lessons and a few supervised pilots, while assistive technology, multilingual learners and career-readiness programs receive exceptions.

No. High schoolers will get AI critical-thinking lessons and supervised pilots. NYC is pausing younger students’ direct use while it studies whether the tools improve learning or simply do the thinking for them.

✨ Meta releases Muse Spark 1.3

🤗 Nvidia buys Hugging Face for $12.9 billion

🚀 Astra nearly maxes out ARC-AGI-3

🌅 OpenAI puts $1 billion behind cyber defenders

⚡ Google launches Gemini 3.8 Flash Cyber

🏗️ G42 considers US ownership to keep AI chips

This week, sit with your child and make sure a strange AI interaction never stays private.

Will splitting one model into public and restricted versions become the new normal? And will other school systems follow New York’s lead? We'd love to hear your thoughts!