Moonshot AI released Kimi K3.1 Pro on October 23 via its apps and API as a post-trained 2.8T-parameter MoE model with 104B active parameters and 1M-token context, followed by open weights on Hugging Face November 4. The model has posted strong agentic and coding results, including top open rankings in Agent Arena and competitive Terminal-Bench 4.0 scores near Claude Opus 5.5, but remains absent from the main Text Arena leaderboard where its predecessor Kimi K3 (Max) sits around 1488 Elo. Open weights typically speed community submissions and blind human evaluations, though broader reasoning gaps versus closed frontier models create uncertainty around high debut thresholds. Trader focus centers on how quickly independent testing translates coding gains into Elo movement.