NaiveAI open-sources high-speed coding model Naive-N0.5-Flash
Understand the story
Chinese company NaiveAI has open-sourced the high-speed coding model Naive-N0.5-Flash. According to an evaluation published by Latent Space on November 11, 2026, the model is built on Xiaomi's MiMo-V2.5 (a 310B-parameter MoE model) and achieved a measured peak speed of about 2,100 tokens per second. The evaluation also examined the model's actual performance on coding tasks. The model is now open-sourced, and related evaluations show it can complete coding tasks while maintaining extremely high inference speed.
AI overview of reporting · Updated 1 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Latent SpaceNaiveAI Promises 2,100 Tokens per Second. How Well Does It Code?
Latent Space reviewed Naive-N0.5-Flash, an open-source model from Chinese company NaiveAI, which is based on Xiaomi's MiMo-V2.5, a 310B-parameter MoE.
Story heat trend
Not enough continuous observations to show a trend yet.