NVIDIA: Nemotron 3.5 Embed 8B (free)

{"about":"### Meaning-based search in a knowledge base\n\nNemotron 3.5 Embed 8B is an NVIDIA model for converting queries and documents into embeddings. These numerical representations help match texts by content when a question uses different wording from the source. The model suits the retrieval layer of a corporate knowledge base, a reference service or an application that selects documents before preparing an answer.\n\nThe underlying NVIDIA model combines text and document images in a shared representation space; generation 3.5 adds retrieval based on the visual content of pages. This Ailoot API listing uses its text workflow: text input and vector output.\n\nAiloot API provides text input and vector output through embeddings. The application receives a representation of the source material and uses it to find similar entries in an index. If a coherent answer with an explanation is needed, the retrieved texts can be passed to a separate generation model together with the user’s question.\n\n### What distinguishes the 8B version\n\nText encoding is based on Nemotron 3 Embed 8B. When creating generation 3.5, NVIDIA retained this text component and its embedding space while extending the underlying model with additional capabilities. The published model card describes an output of 4,096 values, produced by averaging token representations and applying L2 normalization. Match your vector index’s dimensionality to the method’s actual response.\n\nText retrieval was evaluated in 34 languages, including Russian. For multilingual instructions and reference collections, this makes the model a possible shared retrieval layer. The selection quality depends on the database’s contents, document chunking and typical user questions; test it on material from your subject area.\n\n### Workflow\n\n1. Split the source documents into chunks, each containing a self-contained answer or explanation.\n2. Obtain embeddings and store them alongside the text, heading and source reference.\n3. Convert the user’s question with the same model and find similar entries.\n4. Show the retrieved passages or use them in RAG to prepare a source-grounded answer.\n\n### Input size\n\nFor this variant, Ailoot API specifies a text context and input limit of 32,768 tokens. Chunks of a long document can be stored as separate entries with a shared reference to the original material. The model’s available output is a text embedding for the application to use in subsequent processing.","alias_of":null,"aliases":[],"author":"NVIDIA","canonical_model":"nvidia/nemotron-3.5-embed-8b:free","ctx":32768,"formats":["embeddings"],"id":"nvidia/nemotron-3.5-embed-8b:free","input":["text"],"kind":"embedding","max_out":0,"methods":[{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":32768,"family":"openai","fast_multiplier":1,"flex_multiplier":1,"fmt":"embeddings","http":"POST","img_in":null,"img_out":null,"in":0,"input":["text"],"label":"Эмбеддинги","max_out":0,"max_tool_calls":null,"out":0,"output":["embedding"],"path":"/v1/embeddings","request_model_id":"nvidia/nemotron-3.5-embed-8b:free","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]}],"model_aliases":[{"canonical_model_id":"nvidia/nemotron-3.5-embed-8b:free","request_model_id":"nvidia/nemotron-3.5-embed-8b:free","route_label":"Ultra"}],"name":"NVIDIA: Nemotron 3.5 Embed 8B (free)","price":{"in":0,"in_label":null,"out":null,"out_label":null,"sort_in":0,"sort_out":null,"varies":false},"public_id":"nvidia/nemotron-3.5-embed-8b:free","short":"NVIDIA multilingual text embeddings for semantic document search and RAG."}