Migrating: GPT-5.4 Mini to GPT-5.6 Luna

A budget-tier-to-budget-tier migration, and one where the rate-limit group actually changes even though both models sit at the cheap end of their respective generations.

What actually moves

Context window grows substantially — Luna's window is meaningfully larger than Mini's, a real capability gain, not just a generational refresh. Knowledge cutoff moves forward. Rate-limit group changes too: Luna's group carries a notably higher top-tier ceiling than Mini's, so this migration is a throughput upgrade as a side effect, not just a price and capability one. Luna also introduces the GPT-5.6 cache-write charge that Mini, predating that generation, doesn't carry.

What to check before switching

Price moves upward at every tier — Luna costs more than Mini, though still well under the roster's mid-tier pricing. For a workload where Mini's smaller window was a real constraint, this migration is a straightforward win across window, recency and throughput, worth the price step. For a workload that was specifically chosen for Mini's rock-bottom price and never needed a larger window, it's worth confirming the price increase is justified by something this migration actually changes for that workload, rather than switching by default just because Luna is the newer option.

Fast mode carries over

Both models publish a Fast-mode tier, so that option isn't part of this particular decision either way — whichever you pick, low-latency delivery remains available at a premium.