Use casesSelling across languages
Site search that works whatever language the customer types
Semantic search over your catalogue that matches a Dutch query to an English product description — fast, and it cannot invent products.
- How it works
- Your catalogue text is embedded with BGE-M3, a multilingual embedding model, so a query like 'waterdichte wandelschoenen' can match products whose descriptions are in English or German. A reranker then reorders the top matches by estimated relevance. No text generation happens at query time, so responses are fast and the system can only return items that are actually in your catalogue.
- Data you need
- Clean catalogue text: titles, descriptions and attributes per product. This data exists in virtually every shop, but its quality sets the ceiling on search quality — thin one-line descriptions retrieve poorly in any language. The index must be re-embedded when products change.
- What to expect
- It complements keyword and attribute filters rather than replacing them — exact model-number and SKU searches still need classic matching, so run both side by side. Cross-language quality varies by language pair and by how descriptive your product text is; measure on your own real queries rather than trusting benchmark scores. The reranker improves ordering but can still surface odd results for ambiguous queries, and this is an integration project with your shop platform (Shopify, Magento, WooCommerce, custom), not a plug-in.
- Where people stay involved
- None at query time. A person should review a test set of real queries per market before launch, and periodically check zero-result and odd-result queries afterwards.
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.
European languages
Previous-generation models — the board has not been rerun on the 2026 roster, so read the order of magnitude, not a ranking of today's models. And BLEU scores adequacy, not tone: it cannot tell you whether the register fits your brand. That is judged on your own copy.
Independent measurement · openGPT-X / Fraunhofer IAIS · captured 2026-07-10
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.
- 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.
- 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.
- 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 selling across languages
All 16 →Translate your product catalogue with your own word list locked
Draft catalogue translations into EU languages with your brand, material and category terms applied from your own glossary, ready for review.
Let your support team chat with customers in any EU language
Live translation between agent and customer in chat — the customer writes in their language, your agent reads and replies in their own.
Draft replies to customer emails in the customer's language
Reads a customer email in any language, pulls the relevant passages from your policies, and drafts a reply in their language for an agent to send.
Read all your reviews, in every language, as one weekly digest
Translates and tags every review from every market, then writes a short weekly digest in your language with translated quotes attached.
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.