Models
Every model Olaverse ships, in one place — grouped by task, with guidance on which variant to pick.
| Family |
Task |
Page |
| DiacTag |
Diacritization as per-character tagging — cannot corrupt the input |
DiacTag → |
| DiacNet |
Diacritization via seq2seq (restore accents, tones, special characters) |
DiacNet → |
| LID |
Language detection (5–25 languages) |
Language Detection → |
| OTK-BPE |
Tokenization (Nigerian languages, Swahili, Kinyarwanda) |
Tokenizers → |
| Reranker / Embedder |
Retrieval, RAG, cross-lingual search |
Retrieval → |
| MIST |
General-purpose LLMs (8B–140B), chat titles, question generation |
MIST → |
| LegalPeace |
Legal contract reasoning |
LegalPeace → |
| Prism |
Image upscaling, denoising, steganography |
Prism → |
Full Model Index
| Model |
Task |
Size |
Speed |
Install |
LIDLite5 |
Language ID (5 langs) |
1.1 MB JSON |
0.014 ms |
olaverse |
LIDNeural5 |
Language ID (5 langs) |
484 MB |
13 ms |
olaverse[deeplearning] |
LIDLite25 |
Language ID (25 langs) |
~5-10 MB |
<1 ms |
olaverse[lid] |
LIDNeural25 |
Language ID (25 langs) |
~500 MB |
— |
olaverse[deeplearning] |
LIDNeural5_1 |
Language ID (4 Nigerian langs, no English) |
~120 MB |
— |
olaverse[deeplearning] |
MIST-Mini-8B |
General LLM |
15 GB |
~63 tok/s |
olaverse[deeplearning] |
MIST-1-70B |
General LLM |
132 GB |
~23 tok/s |
hosted or multi-GPU |
MIST-1-140B |
General LLM |
256 GB |
~8 tok/s |
hosted or 2× H200 |
MIST-Mini-8B-Thinking |
Reasoning LLM |
15 GB |
~55 tok/s |
olaverse[deeplearning] |
mist-tg-0.3b |
Chat title generation |
~1.2 GB |
— |
olaverse[deeplearning] |
mist-qg-1.5b |
Question generation (25 langs) |
~3 GB |
— |
olaverse[deeplearning] |
LegalPeace |
Legal reasoning |
7B (4-bit) |
— |
olaverse[legal] |
DiacNet (5 Yoruba/Igbo variants) |
Diacritization |
1 MB – 503 MB |
— |
olaverse / [deeplearning] |
diacnet-1.0 |
Diacritization (10 langs) |
~300 MB |
Slow |
olaverse[deeplearning] |
diacnet-1.1 |
Diacritization (10 langs) |
~1.1 GB |
Slow |
olaverse[deeplearning] |
diactag-1.0 |
Diacritization (10 langs), compliance guaranteed |
150 MB / 38 MB int8 |
244 chars/s CPU |
olaverse[deeplearning] / [onnx] |
OTK-BPE-50k (5 Nigerian variants) |
Tokenization |
~3 MB each |
— |
olaverse |
OTK-BPE (9 Swahili/Kinyarwanda/merged variants) |
Tokenization |
varies |
— |
olaverse |
Reranker (2 sizes) |
Reranking |
23 MB – 150M params |
— |
olaverse[retrieval] |
Embedder |
Sentence embeddings (ha/yo/ig) |
~120 MB |
— |
olaverse[retrieval] |
PrismUpscaler (3 sizes) |
Image super-resolution |
~25K params – small |
— |
olaverse[vision] |
PrismDenoiser |
Image denoising |
Small U-Net |
— |
olaverse[vision] |
PrismSteganography |
Image steganography |
Small U-Net |
— |
olaverse[vision] |
Which model should I use?
Diacritization
| Need |
Model |
| Output must never differ from input except in marks |
diactag-1.0 |
| Best Yoruba, Igbo or Hausa accuracy |
diactag-1.0 |
| CPU-only serving at scale |
diactag-1.0 with onnx=True |
| Per-character confidence / review routing |
diactag-1.0 |
| Vietnamese or Portuguese peak accuracy |
diacnet-1.1 |
| Fast Yoruba with no deep-learning extra |
diacnet-yor-viterbi |
| Igbo with no deep-learning extra |
diacnet-ig |
| Automatic language routing (Yoruba/Igbo) |
Diacritizer(model="auto") |
Language detection
| Need |
Model |
| Nigerian languages + English, zero GPU |
LIDLite5 |
| Nigerian languages + English, best accuracy |
LIDNeural5 |
| 25 languages, CPU-only |
LIDLite25 |
| 25 languages, best short-text accuracy |
LIDNeural25 |
| Nigerian languages only (input never contains English) |
LIDNeural5_1 |
LLMs
| Need |
Model |
| Fast everyday use, single consumer GPU |
MIST(size="8b") |
| Structured, detailed output |
MIST(size="70b") |
| Deepest reasoning |
MIST(size="140b") |
| Step-by-step visible reasoning |
MIST(size="thinking") |
| Contract analysis |
LegalPeace |
API Conventions
Olaverse model classes follow a consistent verb pattern — once you know one model, you know them all:
| Method |
Used by |
Meaning |
predict() / predict_proba() |
LID models |
Classify input, optionally with probabilities |
predict_batch() |
LID models |
Batched classification in one forward pass |
restore() |
Diacritizer |
Restore diacritics to plain text |
encode() / decode() |
Tokenizer, Embedder |
Text ↔ token IDs / embedding vectors |
rank() / score() |
Reranker |
Order or score (query, passage) pairs |
normalize() |
TTSNormalizer, NaijaNormalizer |
Expand numbers/abbreviations for speech |
generate() / chat() |
MIST, LegalPeace |
LLM completion / multi-turn chat |
load() |
neural models |
Explicit one-time weight download (lazy elsewhere) |
Small CPU models (e.g. LIDLite5, Viterbi diacritizers) load instantly at construction; neural models expose an explicit load() so the download happens where you expect it.