Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.
Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.
Try it: https://t.co/hhO6qTawgb 🐡
GPT5.2は未来を1年以上先取りしていますね😅GDPvalは中央値で5時間の専門家の作業で評価しています。つまりこれを7割クリアするということはMeasuring AI Ability to Complete Long Tasksで言うと再来年達成するレベルです。2028年に専門研究レベルを置き換えるという宣言はありえないことではないのかもしれません。
GPT-5.2 is here! Available today in ChatGPT and the API.
It is the smartest generally-available model in the world, and in particular is good at doing real-world knowledge work tasks.
If These Benchmarks Are Real, OpenAI Just Ended the Competition
If these numbers are real, OpenAI isn’t just ahead. They’re vaporizing the competition and leaving nothing but a confused cloud of benchmarks behind. This wouldn’t even be a race anymore. It’d be a victory lap.
An interesting fact. At the beginning of the year, almost all models performed better with agentless frameworks (fixed workflow) than with SWE-agent frameworks (including Claude 3.5 and GPT). But starting around March, model scores using the SWE-agent framework completely surpassed those of the agentless ones, marking the arrival of a new era of agentic coding. It’s hard to imagine such a dramatic shift happening in just a few months, and I feel truly fortunate to have been part of it.🥳