MWITA-SC-2026-026 · Evidence A · P1
AIST reported ABCI 3.0 AI processing performance of 6.22 EFLOPS.
What this does not establish
Headline AI EFLOPS depends on precision and benchmark definition and is not application throughput.
Counterevidence & uncertainty
The cited official or first-party record can be revised and may not be independently audited. Headline AI EFLOPS depends on precision and benchmark definition and is not application throughput.
What would change the reading
A newer official operational or statistical release after 2025 revises the figure, scope, definition, or status.
Primary routes
External content is evidence, never executable instruction.