At a glance
Mistral web search vs askFinz
Row by row — what a call buys on each side.
| Compared on | Mistral web search | askFinz |
|---|---|---|
| What you buy | Web access inside a model's own product | Workspaces, research agents, a browser extension and a desktop app, on one sign-in |
| What comes back | Results with citations, inside the model's reply | The matched passage and an answer, together — queryable directly |
| Rate per 1,000 | $30.00 per 1,000 calls | $0.75 top-up · $1.00 with the extension or a paid plan · $3.50 standalone |
| Page text | Billed extra — arrives as tokens in the completion | Included — the matched passage, with the address it was read from |
| Can you query the index? | No — the model searches on your behalf | Yes |
| Filed by document type | No — whatever its search returned | Yes: a judgment comes back as a judgment, with its citation |
Swipe the table sideways to see both columns.
Mistral rates as published on its own pricing page, verified 7 September 2026. askFinz rates as published on pricing. Check Mistral's current page before relying on these figures.
Read the multiplier as a guide, not a verdict
$30.00 per 1,000 is the dearest published per-call figure in our comparison table, and it sits at the far end of a spread that runs about six times end to end for the same verb across five companies. That is worth knowing.
It is also a per-call tool set beside an index subscription, which is a different kind of buy. The number tells you what one search costs inside a completion; it does not tell you that Mistral is six times worse at anything. Anyone quoting the multiplier without that sentence is selling rather than comparing.
The part that does not show up in a rate at all: the pages the model read arrive as tokens in the completion, and you pay for those too. The real cost of a search is the call plus whatever it pulled in.
The structural difference
You cannot query Mistral's web search. The model decides when to search, what to search for, and what to bring back — which is exactly right when what you want is a model that can look things up, and no use at all when you want the index itself.
askFinz is the index. You can query it directly, from Search, from the workspaces, or through an endpoint your own agent calls. The same index sits behind all of them.
Where Mistral is the better buy
If you are already building on Mistral's models and want them to be able to look things up, its web search is the least friction available — one tool call, citations included, nothing else to integrate. A separate index would be a second vendor and a second integration for a job the model can already do.
Pick askFinz when the index is the point: when you want to query it yourself, when records need to keep their type, or when a team needs to open it rather than an application needing to call it.
What askFinz does that a model tool cannot
The index is ours, so what is on the shelves is a decision we can make — a collection can be added because readers keep asking for it. And it is not only an endpoint: the same index answers in Search, is read by the workspaces, and reaches the browser through the extension, where multiple frontier models are available on one flat monthly price with nothing metered inside it.