维基百科链接竞速:高选项数决策演示
让模型在数百到数千个链接中选择下一步,用游戏方式展示决策速度与约束输出的效果。原帖附有 4 分 04 秒视频。

用户发帖称 Jev 的实际智能水平仅相当于 Qwen3.5-2B,唯一优势是超低延迟,并设计了与 Qwen3.5 0.8B、2B、4B 的基准测试对比,提醒大家不必 FOMO。
Jev's actual intelligence level is actually equivalent to Qwen3.5-2B, and its biggest advantage—or rather, its only advantage—lies in its ultra-low latency. Everyone, no need to FOMO 😂 I've designed a benchmark test comparing Jev to Qwen3.5 0.8B 2B 4B, with a thinking budget