The pricing detail worth reading twice: the launch card is $2/M input and $10/M output with cached input 95% off ($0.10), but that’s introductory. Afterwards it moves to $4/$20 (cache $0.20), and Google hasn’t given an end date, so I’d budget on the later card. The 1M output limit (up from 64K) is the line that changes cost planning: output bills at the top rate, so one trajectory that actually uses the full limit is $10 at launch pricing and $20 later, before any input. Google also hasn’t published the input context size or any Batch/Flex rates yet, where the other current Gemini models get both at half price. And for now it’s Fairwind-only, with no API model ID or price-list row as of Oct 1, so Gemini 3.8 Flash ($0.75/$3.75) is still the newest Gemini a normal project can call.