Skip to content

Commercial Support

Olaverse is open source (Apache 2.0) and free to use in production. For teams that need more than the off-the-shelf models, we offer commercial engagements.

Need custom language support?

The published models cover a specific set of languages and domains. If your product needs something different, we provide:

  • ✅ Fine-tuning — adapt DiacNet, LID, rerankers, or embeddings to your domain and data
  • ✅ Custom datasets — corpus building, cleaning, and annotation for underrepresented languages
  • ✅ Domain adaptation — legal, medical, financial, or product-specific vocabulary
  • ✅ Deployment optimization — quantization, CPU/edge targets, latency budgets
  • ✅ Enterprise integration — embedding Olaverse components into your existing pipeline

Why work with us

  • We built the models. Every model in the SDK — DiacNet, the LID family, OTK-BPE, MIST, Prism — was trained by Olaverse Labs, so custom variants come from the people who know the training recipe.
  • Small-model philosophy. We optimize for deployability: a 1.1 MB language detector and a 7 MB diacritizer run where a 7B model can't.
  • Honest benchmarks. Published numbers are measured, limitations are documented on every model card — you'll know what you're getting before you commit.

Get in touch

Contributing

Prefer to build in the open? We actively welcome contributions in:

  • Datasets — parallel text, diacritized corpora, evaluation sets
  • Language support — new languages for LID, DiacNet, and tokenizers
  • Evaluation — benchmark runs, error analysis, per-language reports
  • Documentation — examples, tutorials, translations

See CONTRIBUTING.md to get started.