Link Search Menu Expand Document
AI Alliance GitHub Organization

Open Models and Data Projects

Since the founding of the AI Alliance, many open-weight models have emerged, but the need remains for domain- and use case-specific open models. Build these models requires open, responsibly-governed data sets, addressing concerns for sovereignty and governance.

Several work groups have collaborated on domains such as geospatial, industrial, molecular discovery, and time series. In May 2026, we started Project Tapestry with a broad mission to meet open model and data sets needs that work for all of humanity and all applications.

Project Tapestry

Project Tapestry is a global initiative to build and adapt foundation models with full and flexible support for sovereignty concerns and responsible use of data.

Links Description
Project Tapestry: AI Alliance Page, Technical Website, GitHub Repo AI Alliance icon1
The AI Alliance launched Project Tapestry to build a collaborative foundation for open and sovereign AI. Project Tapestry is an open-source platform designed to enable globally federated development of frontier open models while preserving sovereignty, local control, long-term independence, and the ability to tune models for local cultural and industrial requirements. Tapestry fully embraces the spectrum of data restrictions, such as sovereign boundaries, protection of sensitive and proprietary data, etc. Our goal is to maximize the responsible use of such data sets, using them in ways that benefit all, while fully enforcing any restrictions on their use.

1 The icon indicates an Alliance core project.

Projects for Data Management

Good datasets are essential for building good models and applications. The AI Alliance is cataloging fully-open datasets that will feed into Tapestry’s model building work, but we are also exploring the full spectrum of data constraints and how restricted datasets can be responsibly used to the maximal extent consistent with their constraints.

Links Description
The Open, Trusted Data Initiative AI Alliance icon
Open data has clear license for use, across a wide range of topic areas, with clear provenance and governance. OTDI seeks to clarify the criteria for openness and various restrictions that need to be enforced, then catalog the world’s datasets that meet these criteria. Our projects:

The following notable, recommended projects are developed by Alliance members and affiliated organizations.

Links Description
Docling
Docling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem. Docling is a key tool for the project Parsing PDFs to Build AI Datasets for Science, discussed above. (Principal developer: IBM)
SYNTH
SYNTH is a synthetic dataset built from Wikipedia to teach AI models reasoning skills more efficiently than standard web data. Two small models trained on it achieved top benchmark results using 10-50x less data than comparable models. (Principal developer: Pleais)
CCVec - Common Crawl to Vector Stores
Search, analyze, and index Common Crawl data into vector stores for RAG applications, with three interfaces: CLI, Python library, and an MCP server. (Principal developers: Common Crawl Foundation and Meta)

Open Domain-Specific Models

In addition to Project Tapestry, AI Alliance members have built many domain-specific models and supporting tools, including models for industrial AI, molecular discovery, geospatial, and time series applications.

Open Models
Several AI Alliance work groups are collaborating on the development of domain-specific models:
  • Semikong - The world's first open LLM tuned specifically for the semiconductor industry. (Principal developers: Aitomatic, Tokyo Electron Ltd., FPT Software, and The AI Alliance)
  • Llamarine - An LLM tuned specifically for the needs of the maritime shipping industry.
  • Materials and Chemistry work group (Several developers, including IBM Research):
    • smi-ted - SMILES-based Transformer Encoder-Decoder (SMILES-TED) is an encoder-decoder model pre-trained on a curated dataset of 91 million SMILES samples sourced from PubChem, equivalent to 4 billion molecular tokens. SMI-TED supports various complex tasks, including quantum property prediction, with two main variants (289M and 8×289M).
    • selfies-ted - SMI-SSED (SMILES-SSED) is a Mamba-based encoder-decoder model pre-trained on a curated dataset of 91 million SMILES samples, encompassing 4 billion molecular tokens sourced from PubChem. The model is tailored for complex tasks such as quantum property prediction and offers efficient, high-speed inference capabilities.
    • mhg-ged - SELFIES-based Transformer Encoder-Decoder (SELFIES-TED) is an encoder-decoder model based on BART that not only learns molecular representations but also auto-regressively generates molecules. Pre-trained on a dataset of ~1B molecules from PubChem and Zinc-22.
    • smi-ssed - Molecular Hypergraph Grammar with Graph-based Encoder Decoder (MHG-GED) is an autoencoder that combines a GNN-based encoder with a sequential MHG-based decoder. The GNN encodes molecular input to achieve strong predictive performance on molecular graphs, while the MHG decodes structurally valid molecules. Pre-trained on a dataset of ~1.34M molecules curated from PubChem.
Support Tools