# NVIDIA Nemotron Retriever cuts agent search tokens against rival open retrieval stack

> Original title: Every extra search an AI agent makes uses more tokens.

- Source: NVIDIA AI
- Published: 2026-10-30T00:23:51.000Z
- Bad Influence: https://badinfluence.news/items/lxk0ghcn1mztzukjvo2s67dbw
- Original: https://x.com/NVIDIAAI/status/2115962853463396395

## Summary

NVIDIA says every extra search an AI agent makes consumes more tokens, and it compares its Nemotron Retriever with another open-source retrieval stack on the same questions. The comparison covers the evidence each system retrieves, its accuracy, and its token usage. NVIDIA says the open models and recipes are available on Hugging Face.

## Post

Every extra search an AI agent makes uses more tokens.

Watch Nemotron Retriever tackle the same questions as another open-source retrieval stack. Check out the evidence each retrieves, accuracy and token usage.

Get the open models and recipes on @huggingface → https://huggingface.co/blog/nvidia/nemotron-retriever-world-class-agentic-retrieval

[Video](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Famplify_video_thumb%2F2115962720092921857%2Fimg%2FplyOVUvSUBS7c3NA.jpg&mode=thumb&exp=1793577600&sig=1d40999fe36afbbb)