On July 30, 2026, OpenAI announced a major adjustment to API pricing for the lighter and mid-tier models in the GPT-5.6 family. Specifically, GPT-5.6 Luna prices were cut by 80% and GPT-5.6 Terra by 20%. The move is seen as an important step to pass internal efficiency gains directly to users, aiming to lower the barrier to high-quality AI and address rising cost sensitivity among businesses.

Pricing details

The revised API prices (per million tokens) are as follows:

GPT-5.6 Luna (lightweight, high-efficiency tier): input reduced from $1 to $0.20, output from $6 to $1.20 (an 80% cut)
GPT-5.6 Terra (balanced tier): input from $2.50 to $2.00, output from $15 to $12.00 (a 20% cut)
GPT-5.6 Sol (flagship tier): remains at $5/$30, with a new Fast mode offering up to 2.5× speed at twice the standard price

The GPT-5.6 series was launched for commercial use on July 9, 2026, initially priced at Sol $5/$30, Terra $2.50/$15 and Luna $1/$6. After this price revision, Luna emerges as a highly cost-effective option among mainstream cutting-edge models.

Relationship to previous Mini/Nano models

Within OpenAI's model evolution, Luna and Terra can be viewed as direct successors and upgrades to the earlier “mini/nano” series:

Luna corresponds to the previous nano positioning: prioritizing the lowest cost and highest throughput, suitable for high-concurrency, large-scale automation, simple Agent subtasks and cost-sensitive applications. With the price cuts, its per-million-token cost now approaches or falls below some prior-generation mini models while substantially exceeding them in intelligence.
Terra corresponds to the prior mid-tier-balanced layer (a transitional version between some “mini” models and full models): it balances capability and cost, suitable for everyday coding, knowledge work and medium-complexity Agent flows. The 20% cut further strengthens its “sweet spot” positioning.
Sol continues to play the previous “full/flagship” role, focusing on the highest intelligence and complex long-horizon tasks.

The company clarified that “5.6” denotes the generation, while Sol/Terra/Luna represent sustainable capability tiers that can iterate independently. In several evaluations, Luna and Terra have demonstrated near- or better cost-effectiveness in Agent workflows at a fraction of the cost of Claude Fable 5.

Background and official statements

OpenAI said the price cuts stem from efficiency improvements in model design and inference workflows, including better token generation efficiency and workflow optimization. The company is passing internal cost savings directly to users. Executives including Sam Altman emphasized that this reflects the long-term goal of “making intelligence richer and more affordable.”

At the same time, the cuts apply across the API, ChatGPT Work and Codex product lines, and some auto-routing and moderation functions are being shifted to the lower-cost Luna, magnifying savings.

Potential impact: reshaping cost structures and market competition

1. Impact on developers and enterprises The 80% cut to Luna will materially change the economics of high-concurrency scenarios. Use cases previously constrained by cost—Agent automation, large-scale content processing, real-time assistance—may now scale broadly. Terra's moderate cut makes medium-complexity tasks easier to scale. Overall, the unit cost of intelligence for enterprise AI budgets should fall.

2. Impact on OpenAI itself By widening the price gap between Luna/Terra and Sol, OpenAI can more precisely steer users to choose models by need, increasing overall usage and revenue elasticity. It also reinf