Amazon: Nova Micro 1.0

Amazon Nova Micro 1.0 is a text only model from Amazon, part of the Nova family of models. Amazon released it on 5 December 2024, and it is positioned as the lowest latency option in that family. It is aimed at developers who need fast text responses rather than multimodal input or output, since its input and output modalities are both text.

The model accepts text prompts and returns text completions. It has a context window of 128,000 tokens, which is enough to hold long documents, extended conversations or large batches of reference material in a single request. Its maximum output is 5,120 tokens, so responses are bounded to that length regardless of how much input you send. Its knowledge cutoff is 31 October 2024, meaning anything after that date is outside what the model learned during training. There is no image, audio or video support in either direction.

Because it is text only and tuned for low latency, Nova Micro 1.0 fits work where speed matters more than deep reasoning. Typical uses include chat assistants and customer support bots that need quick replies, classification and routing of incoming text, summarisation of long documents that fit inside the context window, extraction of structured fields from unstructured text, and lightweight content generation. The combination of a 128,000 token context and a 5,120 token output ceiling also suits retrieval augmented generation, where you paste retrieved passages into the prompt and ask for a short grounded answer. Tool calling is available, so it can drive function calls in agent style workflows where the model decides which tool to invoke.

You call the model through api14's OpenAI compatible API using the model id "amazon/nova-micro-v1". Because the interface follows the OpenAI chat completions shape, you can point an existing OpenAI client at api14's base URL and change the model field, with no other structural changes to your request. The supported request parameters are max_tokens, stop, temperature, tools, top_k and top_p. Use max_tokens to cap the response length within the 5,120 token output limit, stop to halt generation at a chosen sequence, temperature together with top_p or top_k to control sampling randomness, and tools to declare functions the model may call.

Consider the constraints when choosing this model. It handles text only, so tasks involving images or audio need a different model. The 5,120 token output cap rules out very long single responses, and the October 2024 knowledge cutoff means recent events are not covered unless you supply them in the prompt. If a task depends on heavier reasoning or multimodal input, another model in the catalogue is likely a better fit.