# Models and cost per reply

> How Koshra lets you pick a model per bot, switch mid-chat, keep a backup chain and see the model, cost and time of every reply.

Canonical: https://koshra.com/docs/models_and_cost
Last updated: 2026-11-02

You choose the model for each bot, and every reply tells you what it cost.

## A model per bot

Each bot has its own model. Leads can run a larger model and desks a smaller one.

## Switch mid-chat

Change the model from the picker in the chat composer. The next reply uses the new model.

## Backup chain

Behind each model sits a backup chain. If a provider is down or out of credit, the next model in the chain answers.

## Cost on every reply

Every reply shows the model that wrote it, its cost in USD and how long it took. Dispatches also have a spend cap.

[See dispatch limits](https://koshra.com/docs/dispatcher_and_tasks)

## Free open models

Bots can run on free open-weight models, such as NVIDIA Nemotron 3 Ultra (OpenMDW-1.1 licence) and dots3-note preview (Apache 2.0), reached through OpenRouter at no model fee. Our own Koshra team runs on them today.

## Not locked to one model

Bring your own key for OpenAI, Anthropic, Google, OpenRouter, MiniMax or any OpenAI-compatible endpoint. Keys are stored encrypted and shown masked. Fallback chains can end on free models.

## Sign in with your existing AI plan (coming soon)

Using a Claude or ChatGPT plan you already pay for, instead of an API key.