Atman Rathod
Yandex has released the weights for its new Alice AI Foundation 80B-A3B Base model, making it available for developers to build upon. The model is now accessible on the Hugging Face platform under a permissive open-source license, allowing the programming community to integrate and modify the system for their own applications.
Technical Specifications of Alice AI Foundation 80B
The new foundation model features a total of 80 billion parameters. However, it operates with a highly efficient architecture, activating only 3 billion parameters per token during processing. This sparse activation design allows developers to run complex artificial intelligence tasks with significantly lower computational overhead compared to dense models of similar size.
Additionally, the model supports a massive 256K context window. This expanded capacity enables the system to process extensive documents, large codebases, and long-form data inputs within a single prompt, offering programmers greater flexibility for context-heavy applications.
- Releases the Alice AI Foundation 80B-A3B Base model weights on the Hugging Face platform
- Features a total of 80 billion parameters for complex artificial intelligence tasks
- Activates only 3 billion parameters per token to lower computational overhead
- Supports a 256K context window to process extensive documents and large codebases
- Uses the Apache 2.0 license to permit free commercial and non-commercial modifications
Open-Source Availability and Apache 2.0 Licensing
Yandex has released the Alice AI Foundation 80B-A3B Base model under the Apache 2.0 license. This licensing structure permits software engineers and researchers to freely use, modify, and distribute the model for both commercial and non-commercial projects without restrictive barriers.
By hosting the model weights on Hugging Face, the company provides immediate access to the global developer community. Programmers can now download the weights and begin experimenting with the new architecture in their own development environments.
The release is expected to drive further development and experimentation in efficient large language models across the open-source community.
Atman Rathod
