---
id: 1957
title: "Optimizing Performance in Generative AI: Trimming Tokens"
source_domain: hackerone.com
source_url: "https://www.hackerone.com/blog/optimizing-performance-generative-ai-trimming-tokens"
published_at: "2024-01-22T09:57:49Z"
tags: ["ai-security", "llm", "performance", "generative-ai", "optimization", "tokens"]
phying_url: "https://phying.com/item/1957"
lang: en
---

# Optimizing Performance in Generative AI: Trimming Tokens

## Summary

Output tokens are generated sequentially and dominate LLM latency; trimming verbose inputs, capping output length, and using smaller models or parallelism can greatly improve generative AI performance.

## Intelligence Metadata

- **Sources:** hackerone.com
- **Published:** 2024-01-22T09:57:49Z
- **Original:** https://www.hackerone.com/blog/optimizing-performance-generative-ai-trimming-tokens
- **Phying News:** https://phying.com/item/1957