AI Compute Stack for Training and Inference | Dweve Core

Dweve Core is a complete AI compute stack for training and inference across supported CPU, GPU, FPGA and browser paths, with explicit determinism contracts.

Choose the audience that matches your question

The page contains three selectable readings of the same subject.

For consumers

The AI compute stack behind Dweve products, running calculations across ordinary processors and specialised hardware without changing the product above it.

For businesses

A complete AI compute stack for training and inference that keeps model authoring, optimisation, execution, serving and verification inside one operator system.

For engineers

An execution-cell registry covering more than 2,900 operations, 76 categories, 25 format families and twelve backend target families.

Dweve Core in one sentence

Dweve Core is a complete AI compute stack for training and inference that keeps model authoring, training, numerical representation, execution, serving and verification in one machine across supported CPU, GPU, FPGA and browser paths.

  • The implementation registry spans more than 2,900 operations, 76 operation categories and 25 registered numerical format families.
  • Core Native is the broad Rust-native execution foundation. Core on Kera is an optional execution foundation available in selected strategic licence packages.
  • Each supported execution cell carries an explicit determinism contract: pure Core Native, seeded Core Native or pure supported Core on Kera.
  • On supported binary workloads, Core uses 96% less energy than a GPU stack.