GPT-6.1 Terra API pricing and cost calculator

How this is calculated

Begin with the calculator's default request. With 100,000 input tokens and 5,000 output tokens, uncached input costs $0.10 and output costs $0.025, so one call is $0.125. With half the input read from cache it's $0.0775, and 100 calls a month start the chart at $7.75. Coding agents are the heavier case. Say a developer runs 30 sessions in a day, each with 25 turns that send an 80,000 token prompt and write 2,500 tokens, with most of the prompt read from cache. One session costs $0.6025 and the day costs $18.075. Without cache hits that day costs $69.375. A cache read is 95% cheaper than uncached input, so keeping the start of every prompt identical is the cheapest habit to build. Writing work has the opposite shape, with small prompts and many calls. A pipeline that drafts 20,000 pieces a month from a 3,000 token brief into 1,200 tokens of text costs $180. Large documents test the prompt boundary. A 300,000 token prompt with 5,000 output tokens crosses it only if 272,000 is below 300,000. It does here, so it costs $0.6375 with no cache hits and $0.3525 with half the input cached, both at the higher tier. Split a large document into passes below the line and compare the totals. Batch prices help with the overnight kind of work. An evaluation run of 50,000 requests, each with 6,000 input tokens and 800 output tokens, costs $500 at the normal rates and $250 in batch mode.

Verdict

Terra is worth testing for the everyday work it's meant for, which is coding, agent loops and writing. Run it on your own tasks and record accepted results, turns per task and output tokens, since a lower rate only helps if the result holds up. Its rates sit below GPT-6.1 Sol on both input and output right now, so the question is whether Terra's results are enough for the task. Keep prompts under the long prompt line where the task allows it.