NaiveAI opens a 309B coding model

NaiveAI released Naive-N0.5-Flash on September 27: an open-weight 309B mixture-of-experts model with 15.5B active parameters and a native one-million-token context. Its Hugging Face model card says the weights and inference code use the MIT license and lists announced API pricing of $0.10 input, $0.40 output, and $0.01 cached input per million tokens, although API access is not live yet. For builders, this belongs in a local coding-agent bake-off alongside Octobench’s evidence that harness choice can swing coding results: the weights are available, but a model this large is an experiment until hardware and independent task results fit your stack.