Beta Use AI with privacy gateway: selecta-gate

About this chat

An internal AI chat running as a Cloudflare Worker in Selecta's own Cloudflare tenant. Every request passes through the Cloudflare AI Gateway (selecta-gate), where it is centrally logged, rate-limited, and available for analysis — the line under each answer shows model, time to first token, and whether the gateway was used.

Selecta Chat

Settings

Appearance

Standing instructions

Saved in this browser and sent along with every message in the background — for example: “Always answer in German. Keep answers short. We are a vending-services company.”

Data

Chats live in this browser and, when you are signed in, in Selecta's own storage.

Compare models

The same prompt, answered by two models side by side. Not saved to the chat.

Answer B —
Answer A · current
Answer B

Pick a model and press Run.

Prompt library

Save prompts you use often and drop them into the message box with one click.

Add a prompt

Hello. The right model saves costs.

  • Llama 3.3 70B (default) — everyday questions, drafting, summaries. Low cost.
  • Llama 4 Scout — quick, simple asks and rewrites. The fastest and cheapest choice.
  • GPT-OSS 120B — tricky, multi-step problems. Thinks before answering, still low cost.
  • Kimi K2.6 / Qwen Coder — top open quality for demanding texts, and code answers. Still low cost.
  • GPT-4.1 / Claude Sonnet 5 — only when top quality really matters. Commercial models — highest cost.