Airbender achieves high throughput on a single GPU using a five-stage proving pipeline, which includes: witness commitment, memory validation, constraint encoding, DEEP polynomial construction and FRI-based proximity proof generation

Quotes that support claims

Airbender’s performance, including the breakthrough to proving chains with a single GPU, is made possible by a set of carefully engineered components. The proving pipeline follows a five-stage structure optimized for both flexibility and throughput: Witness Commitment: Computes low-degree extensions (LDEs) and trace commitments. Lookup & Memory Argument: Validates memory operations using lookup tables. STARK Quotient Polynomial: Encodes circuit constraints via AIR polynomials. DEEP Polynomial Construction: implements FRI batching to reduce proof size. FRI IOPP: Generates the final proximity proof.