Libreagi offers four-times seeding credit for single-copy open-source models

Libreagi is operating a model-saving program that pays seeders of open-source AI weights by scarcity. Its public site says a model with only one copy earns four times the base credit rate, while credit also grows with file size and time. The program accepts weights in safetensors, GGUF and ONNX formats under licenses including Apache-2.0, MIT and the Qwen Community License.

The site describes verification as a chain of custody: each file is checked against its source commit, and staff review every model before it enters the catalog. It also says a signed public catalog lists every model, hash and source. The hash is a cryptographic fingerprint used to confirm that a downloaded file matches the version that was checked. The site gives an IRC channel for questions before application, but does not list a launch date or current model count.

The design treats distribution as infrastructure, not a side effect of popularity. Major model hubs usually rank and recommend models by downloads, which can make older or niche weights hard to find after commercial priorities shift. Libreagi's scarcity multiplier is a direct counterweight: it rewards the rarest copies, the ones most exposed to deprecation. The signed catalog and rebuildable swarm language point to an archive that assumes the hosting site may disappear.

For researchers, small labs and developers, the stakes are access to weights that are still legally redistributable but no longer promoted. A single-copy model can vanish from public mirrors, breaking reproducible experiments, local deployments and fine-tuning pipelines. Libreagi's program does not remove licensing obligations, but it creates a financial and technical reason to keep rare open-source files alive.