A newly released artificial intelligence system can now process up to 1 million tokens of context in a single prompt. Developed by Aleph Alpha, the Kolibri model contains 78B parameters, with 3.46B of those active, and is available for anyone to run on their own hardware. The company distributed the model weights under the Apache 2.0 license, removing the need for cloud-based API reliance.

The European developer announced the release on October 3 and specified that the model was built in Europe. The use of an open-weights license allows businesses and developers to host the model on private infrastructure and maintain full control over their data and processing environment.

Sign in to suggest edits

Key sources

  1. SOURCE@aleph__alpha“78B parameters. 3.46B active. Up to 1M tokens of context. Built in Europe.”x.com
  2. SOURCEhuggingnewshuggingnews.com
  3. DISCUSSIONbastitxnews.ycombinator.com
  4. SOURCEhnaleph-alpha.com
  5. SOURCEmarktechpost.com
Markdown