# How we select and route models in Chota

Canonical HTML: [https://agents.chota.by/en/blog/new-models-may-2026](<https://agents.chota.by/en/blog/new-models-may-2026>)

Language: en

Product Development

Published: 2026-05-11; modified: 2026-07-22.

Author: [Chota Team](<https://agents.chota.by/en/about>) — Product editorial team

This article was prepared by the Chota team. Product claims were checked against the platform's current capabilities, while external facts were checked against the primary sources linked on the page. Material changes are reflected in the update date.

Historical post: capabilities and terms reflect the publication date. The current voice pilot is connected with the team.

[Current product](<https://agents.chota.by/en/product>)

[Current pricing](<https://agents.chota.by/en/pricing>)

Provider catalogs change quickly. We show a model as available only after checking its official status and its presence in the working Chota product.

![Chota request routing across fast replies, document analysis, multimodal processing, and reasoning paths](<https://agents.chota.by/images/blog/model-routing.webp>)

## Why one model is not enough

A quick answer to a common question, a long-document review, and a multi-step integration action have different requirements. The largest model is not always the best business choice: it can be slower and more expensive where a simpler route is sufficient.

We evaluate a scenario across four dimensions: accuracy on real examples, response latency, supported context, and predictable usage cost.

## How we confirm availability

First we check the name, status, and limitations in the provider's official documentation. We then confirm that the model is present in Chota's working configuration and passes our scenario tests. Only after both checks can it be shown to a customer as an available option.

For this update we removed the previous list of specific models because the website must not get ahead of the actual product configuration. The current selection is shown inside the Chota workspace during setup.

## What we test before enabling a model

For customer conversations we repeat the same tests: a question grounded in the knowledge base, an ambiguous request, refusal to invent a fact, correct human handoff, and operation of a connected action. A model change is not an improvement if it only wins on a showcase example.

- Actual availability from the provider and inside Chota.
- Quality on the business's scenarios rather than an abstract leaderboard.
- Latency, cost, and repeatability across requests.
- Safe refusal and human handoff when evidence is insufficient.

## Sources

- [Google — Gemini API models](<https://ai.google.dev/gemini-api/docs/models>)

- [Anthropic — Claude model overview](<https://docs.anthropic.com/en/docs/about-claude/models/overview>)

- [Mistral AI — model catalog](<https://docs.mistral.ai/getting-started/models/models_overview/>)

[How we launched the Chota knowledge base and where its limits are](<https://agents.chota.by/en/blog/knowledge-base-launch>)

[Chatbot, Telegram bot, or AI agent: how we choose for a business](<https://agents.chota.by/en/blog/ai-agent-vs-generic-chatbot>)