NaiveAI/Naive-N0.5-Flash — new model trending #30 on Hugging Face
NaiveAI released Naive-N0.5-Flash, a 309B MoE open-weight model with 15.5B active parameters and a native 1M-token context.
NaiveAI released Naive-N0.5-Flash, an open-weight 309B mixture-of-experts model with 15.5B active parameters, aimed at coding and AI R&D. It uses a native 1M-token context built from 39 sliding-window attention layers (128-token window) and 9 DeepSeek Sparse Attention layers that attend to the top 2,048 tokens, with no full-attention layers. The model continues from the MiMo-V2.5 base and was trained on 3.25T tokens. Weights and inference code are MIT-licensed; NaiveRT claims up to 2,000 tokens/s, and API pricing is $0.10/$0.40/$0.01 per million tokens for input, output, and cache reads.
- 309B MoE with 15.5B active parameters, MIT open weights.
- Native 1M-token context via hybrid SWA and DeepSeek Sparse Attention, no full attention.
- Trained on 3.25T tokens after adapting the MiMo-V2.5 base.
- NaiveRT claims up to 2,000 tokens/s; API priced at $0.10/$0.40/$0.01 per million tokens.
- Vendors
- NaiveAI
- Products
- Naive-N0.5-FlashNaiveRT
- Organizations
- NaiveAI
- AI models
- Naive-N0.5-FlashMiMo-V2.5
From the original article
5 min read at huggingface.co<img class="dark:hidden" src="assets/naive-n0. 5-flash-logo. svg" width="440" alt="Naive-N0.
By NaiveAI · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.