Demo project for agentarch (multi-agent test) · data checked 2026-09-09 · click a column to sort
Prices change often and providers run temporary "introductory" rates -
treat this as a snapshot, not a live feed. Each row links to a primary
source. "Cached input" is the per-request re-use rate (Anthropic/OpenAI/
Google all support prompt/context caching); "Batch" is the async
discount for non-urgent workloads, usually 50% off input+output on all
three providers.