## The problem

On a proxy that routes many models through a single aggregating provider, every row in the models table shows the **same** logo — the provider's. That makes the list visually uniform and hard to scan, even though the models come from very different families.

Concretely, with ~20 Scaleway models configured, every row shows the Scaleway mark, whether the model is Mistral Small, Qwen3, Gemma, Llama 3.3, GPT-OSS, GLM or Whisper. The same happens with any aggregator: OpenRouter, Together, Bedrock, SageMaker, Groq, Fireworks, DeepInfra, Novita.

Scaleway's own console solves this by showing the **model family** icon rather than its own: Mistral rows get the Mistral mark, Qwen rows the Qwen mark, Gemma the Google mark, and so on. It makes a twenty-row list readable at a glance.

## Why this looks cheap to do

Two things are already in place:

1. **The assets exist.** `ui/litellm-dashboard/public/assets/logos/` already ships `mistral.svg`, `qwen.png`, `google.svg`, `meta_llama.svg`, `openai_small.svg`, `deepseek.svg`, `anthropic.svg` — the families that aggregators mostly serve.

2. **The precedent exists.** `providerLogoMap` already maps some entries to a *family* logo ra…