mist-qg-1.5b
Turn any passage into the questions people would ask about it.
mist-qg-1.5b reads a passage and writes natural, search-style questions that the passage answers, in 25 languages. Use it to build quizzes and FAQs, or to create training data for search and retrieval systems.
The problem it solves
Writing good questions is slow, skilled work, and every learning product, help centre and search team needs thousands of them. Large hosted models can do it, but at a per-token cost, with your content leaving your servers, and with weak results in many African languages.
mist-qg-1.5b is small enough to run on one modest GPU and was trained only on questions that proved they could find their own source passage, so what it writes stays grounded in the text.
Highlights
25 languages
From English, Spanish and Japanese to Yorùbá, Hausa, Swahili, Zulu and Amharic. Questions come back in the language of the passage.
Grounded questions
On 625 unseen passages, 95.5% of its questions retrieved their own source passage back.
Clean JSON out
Returns {"questions": [...]}, ready to drop into a database or a training set.
Two jobs in one
A question-generation endpoint for products, and a data factory for (query, passage) pairs to fine-tune retrievers and rerankers.
Run it in a few lines
from olaverse import MISTQuestionGenerator
qg = MISTQuestionGenerator() # mist-qg-1.5b
qg.load()
passage = "Tides are caused by the gravitational pull of the moon and, to a lesser extent, the sun."
qg.generate(passage, n=3)
# → ['What causes ocean tides?', 'Does the sun affect tides?',
# 'Which has a bigger effect on tides, the moon or the sun?']
qg.generate(yoruba_passage, n=3, language="yor")
qg.generate_batch(passages, n=3, language="English")Best for
- Quiz, comprehension and assessment banks for edtech.
- FAQ and “people also ask” content from documentation you already have.
- Creating (query, passage) pairs to fine-tune retrievers and rerankers.
- Teams that want to keep their content in-house instead of sending it to a hosted API.
Use cases
Question generation
Generate natural, search-style questions from any passage across 25 languages, for quizzes, reading comprehension, help-centre FAQs, and assessment banks.
/ ML engineeringRetrieval training data
Generate verified (query, passage) pairs from your own corpus in 25 languages, and fine-tune retrievers and rerankers on data that matches your domain.
More production models
All modelsPurpleMIST-Flash-1.0
typed-decisions 0.720 · African TD 0.668
PurpleMISTPurpleMIST-Mini-1.0
typed-decisions 0.601 · 1.9B
MIST-ExtractMIST-Extract-1.0
91.8% docs fully right · 4B
MIST-ExtractMIST-Extract-Flash-1.0
9.7 docs/s · 2B
LID-608lid-neural-608
0.867 acc · 140M
LID-608lid-lite-608
37 MB · 6,800 texts/s
DiacNetdiacnet-2.0
DER 0.0117 · 582M
DiacNetdiacnet-mini-2.0
DER 0.0141 · 300M
DiacNetdiactag-2.0
DER 0.0130 · 37.9M