This was a week where frontier labs cut prices on their flagship models even as they pushed AI further into higher-stakes territory: scientific discovery, spacecraft control, and the halls of the United Nations.
Anthropic released Claude Opus 5.5, the first model in its new 5.5 family, matching the performance of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5. The model leads on agentic coding and computer-use benchmarks and was tested pre-release by external evaluators including Frontier Design and METR; it's live now via the Claude API and across AWS, Google Cloud, and Microsoft Azure at $4 per million input tokens and $20 per million output tokens. OpenAI answered in kind with updated GPT-6 Sol and Luna models, priced at half the cost of the prior 5.6 series — Sol aimed at complex work like coding, Luna at clerical tasks — with the pair rolling out through ChatGPT Work, Codex, and free and desktop tiers. OpenAI says both hit "Astra-level reliability" with improved factual accuracy and fewer coding mistakes.
The more striking story came from Anthropic's AI biology lab, where Claude identified a previously unknown enzyme system bearing CRISPR-like repeat structures. The company hasn't disclosed what the system might do, but the framing matters: this is being pitched as an example of a model contributing an original scientific finding, not just accelerating a human researcher's existing work.
Elsewhere, the week offered smaller but telling signals of AI moving into domains that used to require constant human oversight. Meta's Muse agent surged to the top of the App Store charts on the back of new third-party Connectors (Notion, Granola) and Stripe-powered in-app payments, sharpening the consumer-agent contest with ChatGPT, Grok, and Claude — though Amazon has since blocked Muse from shopping on its site over notification and disclosure concerns. Space-mining startup AstroForge built an in-house transformer-based system, "Solo," to fly its spacecraft without Earth-based controllers, starting in shadow mode next mission before flying operationally in 2027. Anthropic also published a case study on Claude supporting Ebola outbreak response, and Sam Altman addressed the UN Security Council pushing for global AI standards, with Dario Amodei joining by video — a sign that AI governance conversations are now happening at the highest diplomatic levels.
To watch next week: whether AstroForge's shadow-mode testing and Anthropic's enzyme discovery are early data points in a broader shift toward AI operating with less human-in-the-loop oversight in safety-critical work.