Kimi besser als Qwen-VI abzuschneiden, würde mich wirklich überraschen. K2-VL würde das hingegen nicht (es ist schließlich viel größer als das größte Qwen-VL), aber das Qwen-Team hat zu diesem Zeitpunkt unglaubliche Expertise angesammelt, verdammte 30B MoE konkurrieren mit Gemini. Ich wünsche ihnen viel Glück.

🚀 Hallo, Kimi K2 Thinking! Das Open-Source Thinking Agent Model ist hier. 🔹 SOTA bei HLE (44,9%) und BrowseComp (60,2%) 🔹 Führt bis zu 200 – 300 aufeinanderfolgende Toolaufrufe ohne menschliches Eingreifen aus 🔹 Hervorragend in der Argumentation, agentischem Suchen und Programmierung 🔹 256K Kontextfenster Als Denkagent entwickelt, markiert K2 Thinking unsere neuesten Bemühungen im Testzeit-Skalieren – sowohl das Skalieren von Denk-Token als auch von Toolaufrufen. K2 Thinking ist jetzt im Chatmodus live, der vollständige agentische Modus kommt bald. Es ist auch über die API zugänglich. 🔌 API ist live: 🔗 Tech-Blog: 🔗 Gewichte & Code:

Last week, our reasoning models took part in the 2025 International Collegiate Programming Contest (ICPC), the world’s premier university-level programming competition. Our system solved all 12 out of 12 problems, a performance that would have placed first in the world (the best human team solved 11 problems). This milestone rounds off an intense 2 months of competition performances by our models: - A second place finish in AtCoder Heuristics World Finals - Gold medal at the International Mathematical Olympiad - Gold medal at the International Olympiad in Informatics - And now, a gold medal, first place finish at the ICPC World Finals. I believe these results, coming from a family of general reasoning models rooted in our main research program, are perhaps the clearest benchmark of progress this year. These competitions are great self-contained, time-boxed tests for the ability to discover new ideas. Even before our models were proficient at simple arithmetic, we looked towards these contests as milestones of progress towards transformative artificial intelligence. Our models now rank among the top humans in these domains, when posed with well-specified questions and restricted to ~5 hours. The challenge now is moving to more open-ended problems, and much longer time horizons. This level of reasoning ability, applied over months and years to problems that really matter, is what we’re after - automating scientific discovery. This rapid progress also underscores the importance of safety & alignment research. We still need more understanding of the alignment properties of long-running reasoning models; in particular, I recommend reviewing the fascinating findings from the study of scheming in reasoning models that we released today ( Congratulations to my teammates that poured their hearts into getting these competition results, and to everyone contributing to the underlying fundamental research that enables them!