Celestia and ZKsync reported separate infrastructure tests, putting numbers on data availability throughput and the time needed to produce proofs for Ethereum blocks across two systems.
In a 143-second Fibre run, 120 validators averaged 3.07 Tb/s, according to Celestia’s Oct. 1 announcement. The experimental branch used 2 GiB blobs and allowed up to 2,000 submissions per block.
Those settings differ from the mainnet parameters Celestia described. The company gave 128 MiB as the planned starting blob size and said 200 submissions per block was the current limit. It also said the benchmark’s settings would not be part of the initial mainnet release, so the test figures describe an experimental setup rather than the planned starting configuration.
Celestia said a prospective customer with unusually high throughput requirements prompted the test, but did not identify the customer. It also pointed to agent payments and global finance moving onchain as possible sources of future demand. The company cited an earlier partial benchmark that reached 1 terabit per second.
Celestia plans to bring Fibre to mainnet with capacity suited to early demand and expand it as usage grows. Its stated goal is to reach 3 terabits per second and beyond, but it gave no schedule.
In an Oct. 13 announcement, ZKsync reported a separate test of Airbender Prime, its prover for Ethereum blocks. The evaluation used one NVIDIA RTX PRO 6000 Blackwell GPU to process 7,200 consecutive Ethereum mainnet blocks. ZKsync reported that 99% of the proofs finished within 10 seconds, while the slowest finished before the 12-second mark.
The two results measure different tasks and do not provide a direct comparison between the systems. Celestia measured data availability throughput under experimental conditions; ZKsync reported proof-generation times for a block sample.
