AIBRIEF
OpenAI's three GPT 5.6 pricing tiers shift: Luna drops 80%, Terra 20%, and Sol adds a 'Fast' mode running up to 2.5x faster at 2x the price, with no change in intelligence.
OpenAI reset GPT-5.6 pricing on Wednesday, cutting the list price of the "Luna" tier by 80% and the "Terra" tier by 20% effective immediately. The "Sol" tier keeps its price but gains a new "Fast" mode in the API that runs up to 2.5x faster than Standard at 2x the price, with OpenAI saying the speedup comes with no change in intelligence. Fast mode replaces the prior Priority Processing offering.
The company says Luna is now roughly comparable to "frontier-class" — meaning state-of-the-art — models from about a year prior, at about 6 cents on the dollar per task and nearly nine times the speed. On the Agents' Last Exam professional-work benchmark, OpenAI claims Luna outperforms "Fable 5" at an estimated 99% lower cost per task. Both comparisons are vendor self-claims.
The lower Luna and Terra list prices also flow into metered usage inside paid Codex and ChatGPT Work subscriptions, not just API bills.
Practitioners on Hacker News are already debating routing economics under the new tier math. Independent readouts from trackers like Artificial Analysis's GPT-5.6 landing page and its Sol model card are still pending, so the 2.5x Fast-mode number and the 80% Luna cut should be read as OpenAI's own claims until third-party numbers confirm them.