# Which model for what you are building

Canonical page: https://bertrande.com/en/choose/
License: CC-BY-4.0, reuse with attribution and a link.

No value below was produced, estimated or completed by a language model. Every figure was read from a source we name, on the date we give.

1045 offers are ranked here. Say what you are building and the page shows the offers that fit, ordered on a criterion that is named. There is no score of ours and no opinion: each need states what it requires and what it ignores.

| What you are building | What it requires | Ordered by |
|---|---|---|
| A chatbot people talk to | nothing beyond a price | latency |
| Writing or reviewing code | tools | intelligence |
| An agent that calls tools | tools | price |
| Reading images or documents | vision | price |
| Bulk processing, overnight | nothing beyond a price | price |

**A chatbot people talk to.** Answers have to start fast, so time to first token matters more than raw throughput, and a cached system prompt is what makes it affordable.
**Writing or reviewing code.** Long files mean a large context window, and tool use is what turns a model into an assistant that can read your repository.
**An agent that calls tools.** Tool use is not optional here. Availability matters as much as price, and it is on each offer page, host by host, because one model can sit behind more than thirty of them.
**Reading images or documents.** The model has to accept images as input, which its own vendor declares.
**Bulk processing, overnight.** Nothing waits for the answer, so speed is worth nothing and price is worth everything.

An offer that does not publish the figure a need is ordered by is not ranked last: it is absent, and the page says how many were set aside.

Method: https://bertrande.com/en/methodology/
Report an error: https://bertrande.com/en/corrections/