OpenAI Slashes GPT-5.6 Luna Pricing by 80% — What It Means for Builders

OpenAI dropped the price of GPT-5.6 Luna by roughly 80% to about $0.20 per million input tokens and expanded access to free ChatGPT users in early August 2026. For anyone shipping AI features at scale, the economics of "good enough" frontier intelligence just changed.

Key takeaways

• Luna input pricing fell to ~$0.20 / M tokens — an ~80% cut.
• The model rolled out to free users, widening its reach.
• High-volume, cost-sensitive workloads (RAG, classification, agents) benefit most.

Why the price cut matters

Token cost is the single biggest lever in a production AI budget. An 80% cut turns workloads that were borderline-uneconomical — bulk summarization, per-request classification, always-on agents — into no-brainers. If you priced a feature out six months ago on cost alone, it's worth re-running the math.

Who should switch

Teams running frontier models for volume rather than peak reasoning are the clearest winners. Keep your most demanding reasoning on a top-tier model, but route the long tail of cheap, repetitive calls to Luna and watch the bill drop. As always, benchmark on your own eval set before migrating traffic.

Source: AIToolsRecap — AI News August 2026 and LLM-Stats (retrieved Aug 26, 2026).