1/ Meet dots3-note preview — an open-weight multimodal model built for long-horizon agency in real life. 🌍 Try it: studio.dots.ai/?lang=en Api free now: openrouter.ai/dots-studio/do… 280B total / 16B active 512K context Text + vision + audio Created to sustain reasoning, utilize tools, and adapt through feedback over long time horizons. Copy

2/ 🧠 Train agents for tasks that last hours—not minutes. Say hello to TEMPO. By splitting long trajectories into macro-steps, it helps agents stay on track by switching between Actor and Critic—assessing progress and course-correcting before the task concludes. During a hidden-rule game, two distinct branches secured the exact same reward after 64 rounds. Yet the Critic caught what the raw reward missed entirely: Branch B misunderstood the objective and kept going in the wrong direction. Branch A found the underlying conflict rule and moved toward a feasible solution. Branch A: 3.8 | Branch B: 2.29 Equivalent reward. Very different progress.

4/ 🔁 Learn in unfamiliar environments. Through interaction and feedback, dots3-note preview can: Explore → Form & test hypotheses → Update memory → Reuse what it learns This adaptive capability shines in extended interactive scenarios like ARC-AGI-3.