Category

Ornith-1.5-122B

Uses reinforcement learning to improve task generation, solving strategies, and solution rollouts in a continuous loop. This 122B MoE model targets coding, reasoning, and agent tasks, supports a 262K context window, and can run on one 80GB GPU.

Viggle Remake

Remakes existing footage from text instructions, changing characters, props, or scenes while preserving the original motion, camera, cuts, and sound. Use it in the Viggle app or run it on your own GPUs.

Mage-VL-Cowork

Combines screen-based interaction and terminal execution in one 4B vision-language policy, choosing between GUI actions and direct file operations by task. Scores 66.0% on 361 OSWorld-Verified tasks; shell access makes sandboxing essential.

LongCat-2.5

Handles long-horizon agent tasks across codebase development, project builds, and visual work. It combines native visual tokens with large-scale agentic reinforcement learning and supports a one-million-token context.

Codex Tool-Use Trajectories

Collects multi-step tool-use sessions from Codex Desktop, including 437 trajectories and 314 subagent delegation examples. The work is concentrated on Windows-based C#/C++ development and video-codec reverse engineering.

Mem0

Memory layer that lets AI agents remember users and context across sessions.