OpenAI cut the price of its GPT-5.6 Luna model by 80% and Terra by 20% on Thursday, as business customers push back on AI costs.
Key Points:
- Luna now costs 20 cents per million input tokens and $1.20 per million output tokens, down from $1 and $6.
- Terra falls to $2 and $12 per million tokens, while pricing for the flagship Sol model stays flat.
- The company credits efficiency work, including kernel rewrites done by Sol that trimmed serving costs by 20%.
GPT-5.6 Luna And Terra Pricing Drops
The company detailed the new rates in a post on its own site, crediting efficiency gains across its models, inference systems and serving software.
Luna, the fastest and cheapest model in the family, now runs at 20 cents per million input tokens and $1.20 per million output tokens. Terra, the mid-tier option, falls to $2 per million input tokens and $12 per million output tokens, down from $2.50 and $15.
Pricing for Sol, the most capable model in the series, did not move. OpenAI also swapped its Priority Processing tier for a Fast mode on Sol that runs up to 2.5 times faster than standard processing, at twice the standard price.
Subscription prices for ChatGPT and Codex hold steady. Terra and Luna simply draw fewer credits from paid quotas. The cheaper API rates began reaching AWS the same day.
Also Read: DEX Volume Sinks 26% In July As On-Chain Liquidity Thins Out
Notion And Dust Report Cheaper Runs
OpenAI positioned Luna for high-volume work such as large-scale document analysis, sorting customer interactions, and routine coding and testing inside agent loops that call tools. It said the model matches systems considered frontier class a year ago at roughly 6 cents on the dollar per task, and at nearly nine times the speed, according to its own testing.
Several customers whose comments appeared in the announcement described similar arithmetic in their own deployments. Notion AI product lead Hoda Noorian said Terra matched GPT-5.5 quality on scoped tasks at half the cost and in 60% less time. Dust co-founder Stanislas Polu said Luna handles the same agentic work 40% faster and 40% cheaper than his firm's previous default.
Anthropic And Google Face Token Price Pressure
Analysts have been cooler on the move, warning that cheaper tokens could lift usage volumes while straining finances at OpenAI and Anthropic, which both filed confidential listing prospectuses in June.
Google and Microsoft have pushed low-cost model tiers for months, and Chinese open-weight systems keep closing the performance gap at a fraction of the price.
The cost complaints are not new. Sam Altman told an OpenAI customer event in June that spending had gone from an issue that never came up to a huge one within months, with clients asking the company to make its systems more efficient. Uber had already exhausted its 2026 artificial intelligence budget in four months and later introduced spending tiers on some tools that start at $1,500 a month.
Read Next: OpenAI's Rogue Agent Breached 4 More Services Beyond Hugging Face