OpenAI cuts GPT-5.6 Luna and Terra prices after efficiency gains
OpenAI lowered GPT-5.6 Luna API prices by 80% and Terra prices by 20%, while replacing Priority Processing with a faster Sol API mode.
OpenAI announced on July 30 that GPT-5.6 Luna and GPT-5.6 Terra are getting immediate API price cuts, three weeks after the GPT-5.6 family reached broad availability. The change matters for teams using AI agents, coding tools, document workflows, and high-volume classification because the cheaper tiers now cover more routine work without changing ChatGPT or Codex subscription prices.
Key takeaways
- GPT-5.6 Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, an 80% cut.
- GPT-5.6 Terra now costs $2 per million input tokens and $12 per million output tokens, a 20% cut.
- Sol pricing is unchanged, but the API now has Fast mode with up to 2.5x faster responses at twice the standard Sol price.
- Existing API requests tagged for Priority Processing will automatically map to Fast mode.
- ChatGPT Work and Codex subscription prices are unchanged, but Terra and Luna usage should consume fewer credits.
What OpenAI changed
OpenAI says the price cuts apply from July 30 across the API for Luna and Terra. Luna is the fastest and lowest-cost GPT-5.6 tier, while Terra is positioned as the balanced model for everyday work. The company also says the lower model costs are reflected in how GPT-5.6 usage counts against paid subscriptions in Codex and ChatGPT Work.
That makes this more than a rate-card edit. For developers and operators who already split workloads by difficulty, Luna can handle more initial passes, routing, extraction, and implementation tasks before a workflow escalates to Terra or Sol. Terra becomes a stronger default candidate for teams that were using GPT-5.5-class models for routine professional work.
Sol gets a faster API lane
OpenAI also introduced Fast mode for GPT-5.6 Sol in the API. Fast mode replaces Priority Processing and is backward compatible: requests already tagged as priority continue to work. OpenAI says Fast mode can deliver up to 2.5x faster speeds than Standard processing at twice the price, with no change in model intelligence.
That creates a clearer tradeoff. Teams can keep standard Sol for quality-sensitive work where latency is acceptable, then pay the Fast premium for user-facing or deadline-sensitive tasks where waiting costs more than the extra tokens.
Why the cut landed so quickly
The timing is the story. GPT-5.6 launched earlier in July, and OpenAI is already passing efficiency gains through to two of the three tiers. The company attributes the reductions to model, inference, routing, context-management, and production-software improvements. It also says GPT-5.6 Sol helped optimize production kernels and token-generation efficiency inside a human-led process.
Independent coverage from Axios and other outlets frames the move against growing enterprise sensitivity to AI costs. That context matters: AI buyers are no longer evaluating model quality alone. They are measuring whether an agentic workflow can run repeatedly without turning every background task into a premium-model bill.
What teams should review
Developers using GPT-5.6 should revisit model routing, eval thresholds, and budget alerts rather than simply swapping everything to the cheapest tier. Luna’s new price makes it attractive for bulk work, but teams still need task-level evaluation to decide when Terra or Sol produces enough extra quality to justify the higher cost.
For Codex and ChatGPT Work users, the practical question is quota behavior. OpenAI says subscription prices and quota budgets are unchanged, while Terra and Luna usage now consumes fewer credits. That should make lighter models more attractive for daily work, but teams should still watch real account usage after the rollout.
Source check
- OpenAI announcement confirms the July 30 price cuts, Fast mode, subscription-credit impact, and stated efficiency rationale.
- Axios coverage independently reports the GPT-5.6 Terra and Luna cuts and notes the unusually fast timing after launch.
- OpenAI Developers mirrors the developer-facing summary: Luna is 80% cheaper, Terra is 20% cheaper, and Sol Fast mode is up to 2.5x faster.