RADIX WikiRADIX Wiki

Overview

Both Radix and Polkadot address scalability through sharding, but with very different architectures:

Sharding Model

Polkadot uses a hub-and-spoke model: parachains (application-specific shards) connect to a relay chain that provides shared security. Parachain slots are limited and auctioned — creating artificial scarcity.

Radix uses Cerberus with effectively unlimited shards. Consensus is braided dynamically per-transaction across relevant shards — no slot auctions, no capacity limits.

Composability

Polkadot's XCM (Cross-Consensus Messaging) enables cross-parachain communication, but it's asynchronous. Radix provides atomic composability across all shards — critical for DeFi where partial execution can mean lost funds.

Scalability in practice (2026)

The braided, effectively-unlimited-shard architecture described above is the design of Cerberus, not the current state of the live network. Babylon mainnet today runs a single, unsharded Cerberus instance; multi-shard consensus is being delivered through the Hyperscale program (the eMunie → Tempo → Cerberus → Xi’an lineage). A January 2026 public test sustained 500,000+ real cross-shard transactions per second (peaking above 800,000) across 128 shards, as detailed in the Radix Hyperscale update.

Against Polkadot, the practical difference is capacity model: Polkadot’s throughput is bounded by a fixed number of auctioned parachain slots on its relay chain, whereas Radix’s Hyperscale target is to add shards without slot auctions or capacity caps, preserving atomic composability across every shard rather than relying on asynchronous cross-parachain messaging (XCM).