Naive-N0.5-Flash

Naive-N0.5-Flash is a language model for writing code and assisting AI research. Naive AI announced it on 27 September 2026 and published its weights under the MIT licence. Downloadable weights let an organisation run the model on its own infrastructure.

Part of the model runs on each token

Naive-N0.5-Flash has 309 billion total parameters, with 15.5 billion active per token. Parameters are the learned numerical settings inside a model; a token is a piece of text. Its mixture-of-experts architecture selects part of the network for each step.

The model card specifies a native one-million-token context. This is the space available for the prompt, supplied documents and conversation. The large total parameter count still creates substantial memory requirements when hosting the weights.

Hosted access has an announced price

Naive’s launch lists planned API prices of $0.10 for input, $0.40 for output and $0.01 for cached input per million tokens. The announcement describes hosted access as forthcoming. Readers should check the provider’s current access page before treating those rates as an available service.

The launch also describes a specialised inference runtime, with code promised by 12 October. Its speed claims are vendor measurements on the stated hardware. A deployment comparison should measure task success and complete request latency as well as generation speed.