Why does attaching a screenshot make your assistant noticeably slower to start answering?

Attempt it

First, prefill grows: a single image adds hundreds to a few thousand tokens that the model must process before emitting the first output token, so time to first token stretches. Mitigate by downscaling client side, using a lower detail setting, or sending a crop of the relevant region. Second, transport and preprocessing overhead: the base64 payload upload plus provider-side resizing and tiling happen before inference even starts. Mitigate by compressing dimensions before upload and reusing cached results for repeated images.

Your answer

Quick reference

TL;DR

Image attachments inflate prefill (hundreds to thousands of vision tokens to process) and add transport plus preprocessing overhead before inference starts. Downscale, use prompt caching, and crop to attack both.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Imagine you ask a friend a question by text. They reply quickly. Now you also send them a big photo of your screen and ask the same question. They have to wait for the photo to download over slow Wi-Fi and then study it before they can answer. Two separate slowdowns: the photo took time to arrive, and reading it took longer than reading your short text. Sending a smaller crop of just the part you care about helps both. If you keep sending the same photo, remembering it from last time helps even more.

Key concepts

Multimodal AI

Detailed answer & concept explanation

~8 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Sign in for 5 free unlocks/day or go Pro for unlimited. See Pro pricing.

Quick scan · 2–4 min

Everything important, quickly.

5 min: latency budget breakdown, transport plus preprocessing cost, prefill expansion, mitigation order, operational metrics.

Real-world examples

Real products, models, and research that use this idea.

Claude Opus 4.7 and Sonnet 4.6 vision both show TTFT lifts of 300 to 800 ms per high-resolution image attachment.
OpenAI GPT-5.5 vision exposes a low detail mode that caps image tokens to a small fixed budget for fast scene-gist tasks.
Gemini 3.1 Pro context caching reduces repeated-image prefill cost dramatically for document-assistant workloads.
Anthropic's prompt caching documentation explicitly calls out image prefixes as the highest-leverage caching target.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QHow much does base64 actually add versus binary upload?

Sign in or go Pro to see the approach for this follow-up.

QWhy does prompt caching speed up prefill, not just reduce cost?

Sign in or go Pro to see the approach for this follow-up.

QHow does multi-image input affect latency?

Sign in or go Pro to see the approach for this follow-up.

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Blaming inference. The latency hit is mostly prefill and preprocessing, both of which happen before the model emits a single output token. Tuning generation parameters does not help here.

Sign in or go Pro to see all red flags and common mistakes. See Pro pricing.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.