NaiveAI 宣称每秒 2100 tokens，编码能力实测如何？
原标题：NaiveAI Promises 2,100 Tokens per Second. How Well Does It Code?
AI 摘要
Latent Space 评测了中国公司 NaiveAI 推出的开源权重模型 Naive-N0.5-Flash，该模型基于小米 MiMo-V2.5 继续训练并适配 DeepSeek 稀疏注意力，主打高速生成（宣称约 2100 tokens/秒）。在 30 道不同难度题目的测试中，其吞吐接近或超过 1000 tokens/秒，中位响应时间 3.1 秒快于 DeepSeek V4.1 Flash 1017 和 GLM 5.3 Flash，但思考模式下平均 token 消耗显著更高、难题上平均耗时反而更慢。官方基准显示其编码能力介于两个对比模型之间，NL2Repo-Bench 得分领先。
正文节选
Editor’s note: With this post, we are launching a new series focused on trying new models and products with unbiased reviews. Reach out with suggestions to tips@latent.space. Naive-N0.5-Flash is a new open-weights model built for AI coding tasks and tuned to generate as many tokens per second as possible. It was developed by a Chinese company called NaiveAI, with the aim of speeding up engineering and research workflows so that teams can experiment, iterate and solve complex problems faster. The