# Out of Sight, Still in Mind: Token Compression for Omni-LLMs

> Source: <https://aiflash.com/news/154142/>
> Published: 2026-11-19 05:30:02+00:00

**Recommendation:** Target visual tokens first when trimming omni-LLM inference costs, since the paper attributes the vast majority of input tokens to the visual stream.
More in the playbook "Controlling LLM cost": <https://wpnews.pro/playbooks/llm-cost>

The goal of this paper is to reduce the input token cost of Omni-modal large language models (Omni-LLMs) at inference time. Omni-LLMs reason jointly over audio, video and text, but the cost of the three streams is highly unbalanced: visual tokens account for the vast majority of the input, and are h
