New Frontier Red Team blog: Phase 2 of Project Fetch, where we test how well Claude can program a robodog.
Opus 4.7, on its own, was ~20x faster than last year's best human team aided by Opus 4.1. (The robodog, alas, still failed to fetch a beach ball.)
https://t.co/CgbBtRf85e
100 STING drones have been procured for Ukrainian defenders, thanks to @TimothyDSnyder 🔥
Already on the frontlines, helping save lives and strengthen the army.
Within the first 7 months of deployment, STINGs helped destroy 3000 russian drones🧵
https://t.co/hBmVGuU8lx
The drone pilots of the motorized infantry battalion of the 118th Separate Mechanized Brigade carried out a unique mission to save these adorable cuties🐱
Using a large drone, they successfully evacuated a mother cat and her 5 kittens🧵
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
AI can give researchers the freedom to pursue “crazier” ideas.
For Terence Tao, AI creates more room to experiment, test unexpected paths, and discover what might otherwise stay out of reach.
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
❤️🩹 Pasha spent three days alone with injured legs before his brothers-in-arms found him.
Through years of painful recovery, he managed to come out the other end stronger.
@thsottiaux Yes! I often do tasks that are not time-critical and would not mind waiting a bit longer if I save money. Perhaps we could have all kinds of options, from "charge more and give me the result ASAP" to " save money, I won't mind waiting till tomorrow".
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.
For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.
An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.
This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
Introducing Gemini 3.5: our newest family of models combining frontier intelligence with real-world action.
The first release is 3.5 Flash, our strongest model yet for agents and coding 🧵
Introducing Daybreak: frontier AI for cyber defenders.
Daybreak brings together the most capable OpenAI models, Codex, and our security partners to accelerate cyber defense and continuously secure software.
A step toward a future where security teams can move at the speed defense demands.
Based on the results from April, our long-range sanctions have reached a new level across three components: reducing Russia’s oil revenues, as well as the range and intensity of sanctions. It is important that not only is the target itself reached, as defined by the operational objective, but that the downtime of the target is increased or, at the very least, its operations are significantly reduced.
According to the most conservative estimates, since the beginning of the year, the aggressor state has lost at least $7 billion solely as a direct result of our precise sanctions against Russia’s oil industry and refining sector – due to direct hits, downtime, and delays in shipments.
I am grateful to all our warriors of the Defense Forces of Ukraine who, together with the Security Service of Ukraine and our intelligence agencies, are delivering these results. We will scale up our long-range systems capabilities. Decisions are being prepared.
Glory to Ukraine!
11 enemy targets downed by interceptor STING 🚀
All Shahed and Gerbera interceptions were carried out by the SBU “Alpha” unit — one of the most effective units in countering Shahed-type attack UAVs in 2025.
Introducing GPT-5.5
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex.
Introducing Claude Opus 4.7, our most capable Opus model yet.
It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back.
You can hand off your hardest work with less supervision.