Base44 Moves Scheduled Automations to GPT-5.6 Luna
Base44Base44 switched scheduled, entity-triggered, and workflow automation runs from Haiku 4.5 to GPT-5.6 Luna, OpenAI's lightweight tier in the GPT-5.6 family. The change is available directly in the Superagent model picker, and automation credit costs remain unchanged despite the model swap.
A New Default Engine for Background Automations
Base44 updated the model powering its automation system, scheduled tasks, entity-triggered actions, and workflow runs, switching the default from Anthropic's Haiku 4.5 to GPT-5.6 Luna, OpenAI's cost-efficient tier within the GPT-5.6 family.
What Changed for Builders
Automations that previously ran silently in the background on Haiku 4.5 now execute on GPT-5.6 Luna instead. The switch is visible and configurable directly from the Superagent model picker, so builders who want to override the default can still select a different model for their automation runs. Importantly, credit costs for automation runs are unchanged, meaning the model swap does not carry a pricing implication for existing workflows.
Why It Matters
Background automations are a workhorse feature for many Base44 apps, powering scheduled reports, data-driven triggers, and multi-step workflows that run without a user actively watching. Moving the default engine to a newer, OpenAI-backed lightweight model suggests Base44 found meaningfully better quality or reliability from Luna at comparable cost and latency to the previous Haiku 4.5 default, a quiet but consequential change for anyone relying on unattended automation runs.