Pleias
Open training data and small models, now with enterprise data tools
Overview
Pleias is a Paris company that publishes Common Corpus, a 2.27-trillion-token dataset of public-domain and permissively licensed text, and small Apache-2.0 models trained on open or synthetic data. It now also sells two beta products: Synth for synthetic training data and Stratum for on-premise document processing.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Pleias
- Common Corpus: a 2.27-trillion-token open pretraining dataset with provenance
- Pleias 1.0 base models (350M, 1.2B, 3B) in eight European languages
- Pleias-RAG-350M and -1B with built-in source citation
- The SYNTH synthetic dataset and Baguettotron models, including a mixture-of-experts variant
- CommonLingua, a 334-language identification model
- Synth and Stratum enterprise data tools, both in beta
Pleias' durable contribution is Common Corpus, a 2.27-trillion-token pretraining dataset built only from public-domain and permissively licensed text and accepted as an oral at ICLR 2026. Its small Apache-2.0 models suit teams that need documentable provenance, though benchmark claims come from Pleias' own papers.
The company has moved toward enterprise data tooling: Synth and Stratum are in beta with no published prices. No model supports Southeast Asian languages; the dataset, which added Southeast Asian sources in 2026, is the more relevant release for the region.
What we found
Pleias is a Paris company, registered as Pleias SAS in December 2023, that began as a small-model lab and now describes itself as building the data infrastructure layer for enterprise agentic AI. It has no app-store or review-site presence, so the signals worth reading are its releases, peer review and named deployments.
The strongest is Common Corpus, an open pretraining dataset of public-domain and permissively licensed text that now stands at about 2.27 trillion tokens. Its paper was accepted as an oral at ICLR 2026, recorded on OpenReview as well as on the company's site. A February 2026 update added more than 270 billion tokens, over half from non-western countries, and Pleias names Southeast Asia among the new sources alongside China, Japan, Korea, India and Africa.
The models built on it are small and Apache-2.0. Pleias 1.0, released in December 2024, offered 350M, 1.2B and 3B base models across eight European languages. Pleias-RAG-350M and -1B followed in April 2025 with built-in citation of sources, and the lab's own paper says they are competitive with Qwen-2.5-7B, Llama-3.1-8B and Gemma-3-4B on multi-hop question answering; we found no independent replication. Newer releases are trained on SYNTH, a synthetic corpus amplified from Wikipedia articles: Baguettotron at 321M parameters and a mixture-of-experts variant with about 1B active parameters.
The commercial side is new and still in beta, contrary to earlier versions of this page, which said Pleias sold nothing. Synth turns a customer's own documents into training data, on-premise or by API, with pay-as-you-go pricing but no published figures. Stratum, a document-parsing and PII-redaction layer, is in closed beta. The one named production deployment is Sillon, a 600M model built for the Paris transit operator RATP.
None of the models targets Southeast Asian languages, so for readers in the region the open dataset is the more useful release, and the paid products are reached through a demo call.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Pleias's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Pleias unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Pleias directly →
Spotted something out of date? Suggest an update →
Alternatives to Pleias
More in LLMs & Chat
