ML Env

Use casesSearch & discovery

Stop losing sales on 'no results found' pages

When keyword search returns nothing, the query falls through to embedding search and shows the closest products instead of a dead end.

How it works
When your existing keyword search returns nothing, the query falls through to embedding search, which returns the semantically closest products in your catalogue. Optionally the small language model rewrites the failed query first — fixing phrasing, expanding abbreviations — before the retry. Zero-result searches are directly countable in your own search log, so you can measure your zero-result rate before and after on your own data rather than trusting anyone's benchmark.
Data you need
Your product catalogue plus your on-site search logs, so you can see which queries currently fail and verify the fix on real examples. If your shop platform does not log searches, turn that on first.
What to expect
It reliably turns most dead-end pages into pages with plausible near-matches, and the effect is measurable on your own logs. But if the shopper wants something you genuinely do not sell, the best possible outcome is a close alternative — which can feel like bait if presented as an exact match, so rescued results should be labelled honestly ('nothing exact, closest matches'). The fallback path adds a little latency. Embeddings handle most EU languages, but spot-check quality on your shop's language and product jargon before trusting it unattended.
Where people stay involved
Someone reviews the top recurring rescued queries weekly at first: the fallback can surface products that are 'closest' but commercially wrong — a premium item for a budget query, for instance.

Which model, and what it costs to run

Qwen3-8B

No public benchmark ranks this particular job, so this is not a leaderboard pick — it is where we would start: small, permissively licensed, and strong wherever it has been measured. We prove it on your own content before anything ships.

Licence
Apache-2.0
Weights at 4-bit
5 GB
Context
32K tokens
Publisher
Alibaba (Qwen Team)

What the hardware costs

One 48 GB card holds it

Rent in the EU
$1.60/hrScaleway, Paris (PAR2)
Buy the card
$7,569new, one-off
Or rent it by the token
$0.04 / $0.04per M in / out · DeepInfra

Hardware only, third-party prices from 2026-07. The figure excludes the KV cache, which grows with context length and how many people use it at once — sized properly in a conversation, not guessed here. Renting by the token is cheaper up front; why our customers still self-host is below.

Search over your own catalogue

No benchmark measures this

MTEB and BEIR rank embedding models, but their evaluation sets sit inside the models' training data. Scores fall by double-digit nDCG on a private corpus, so the ranking does not transfer to your catalogue.

So we do not show a chart here. We measure it on your own content, in the first week, and you see the result before anything ships.

The API is cheaper per token. Here is why our customers don't use it.

We will not pretend otherwise: renting a model by the token from a serverless API costs less per million tokens than a card we run for you. We show that price on every use-case page. What it does not include is the part a shop with a customer database actually pays for.

  1. 01

    Your data never leaves hardware you can point at

    A serverless "we don't retain your data" is a clause in a contract. Running the model on a card in Amsterdam is a fact of architecture: your catalogue, tickets and customer records are never sent to a third party at all. For a GDPR audit, that is the difference between a promise and a floor plan.

  2. 02

    The price cannot move without your say-so

    A serverless rate card is somebody else's lever. The provider can raise the price, retire the model, or change the terms, and your cost moves with it. The same model on the same card costs the same next year — you own the number.

  3. 03

    The model cannot be taken away

    Hosted APIs deprecate models on their own schedule; the one you built on can be gone in a quarter. An open-weight model on your own hardware runs for as long as you keep the lights on. No vendor can end-of-life it out from under you.

And the price gap closes with volume: past a card you keep busy — very roughly four billion tokens a month — owning is cheaper outright, even before the three reasons above.

Getting a case like this one from a conversation to production takes about two months, and you can stop at the end of any phase.

Also in search & discovery

All 12 →

Next

Tell us what your team does by hand

Describe the process that takes the most time. We will say plainly whether a model is the right tool for it.