1T / 63B (MoE)
MIT
World's first open-source trillion-parameter reasoning model based on hybrid linear attention architecture; 1:7 MLA + Lightning Linear Attention ratio reduces memory access overhead by 10x and increases generation throughput by 3x for sequences exceeding 32K tokens; 128K context (expandable to 256K via YaRN); enhanced deep reasoning with dense rewards for rigor feedback during RLVR; large-scale fully-async agentic RL training for long-horizon task execution; IMO 2025 gold medal level (35/42 points); CMO 2025 (105/126 points, exceeding gold threshold of 78); SOTA on Gaia2-search among open-source models; compatible with Claude Code and OpenClaw frameworks; supports agentic programming, multi-step tool calling, cross-application tool collaboration