ML Env

Use casesVoice of customer

Ask questions of your own review pile, in plain language

Type a plain-language question about your own reviews and get an answer with the actual quoted reviews it drew from.

How it works
All reviews and survey verbatims are embedded into a searchable index. You type 'what do customers say about the sizing of the linen range?'; BGE-M3 retrieves the relevant reviews, the reranker orders them, and the large model answers with a summary plus the quoted reviews it drew from. It is a research tool for ad-hoc questions between reports, not a dashboard.
Data you need
A consolidated corpus of reviews and survey verbatims with product and date metadata attached, so answers can be filtered ('only 2026 reviews'). In practice reviews live scattered across Trustpilot, Amazon, Google and shop-platform review apps — pulling them into one place and keeping the index reasonably current is a real setup task, not a given. With under a few hundred reviews in total, reading them directly is often the better tool.
What to expect
It can only answer from what customers wrote — 'why did March sales dip?' is not in the reviews, and the model may nonetheless attempt an answer, so users need to learn what questions the corpus can carry. Retrieval works on meaning rather than exact words, so common paraphrases ('runs narrow' versus 'tight fit') tend to be found, but product-specific slang and niche jargon will sometimes be missed, and retrieval quality is less even across languages in the mixed-language review sets common for EU shops. Answers about rarely-reviewed products rest on a handful of reviews and sound more confident than that.
Where people stay involved
The person asking reads the cited quotes, not just the summary — the citations exist precisely so a human can verify. Answers should never be forwarded to a customer or put in marketing copy without that check.

Which model, and what it costs to run

Qwen3-8B

This job answers while someone waits, so latency comes first: we hold it to a single small card and, within that, take the best independent score on sticking to the document.

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.

Sticking to the document

Vectara Hallucination Leaderboard · HHEM · 17 of 18 models measured

Measured on public documents, by a model acting as judge. Read it beside the answer rate: the lowest hallucination rates on this board belong to models that simply decline more often.

Phi-43.7% · answers 80.7% ⚠Llama 3.3 70B4.1% · answers 99.5%Gemma 3 12B4.4% · answers 97.4%Qwen3-8B4.8% · answers 99.9%Mistral Small 3.25.1% · answers 97.9%Granite 4.0 Small5.2% · answers 100%DeepSeek-V3.25.3% · answers 96.6%Qwen3-14B5.4% · answers 99.9%
Shorter is betterFree to serveConditions apply⚠ answered under 95%

Independent measurement · Vectara · board updated May 11, 2026

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 voice of customer

All 15 →

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.