GPT-5.6 Luna drops 80% and Terra 20%. Cheaper tiers only help if you map work to the right model.
Via OpenAI: Advancing the price-performance frontier with GPT-5.6
On July 31, 2026, OpenAI said it is passing GPT-5.6 efficiency gains to customers with lower prices: Luna, the fastest and most affordable tier, costs 80% less, and Terra, the balanced everyday-work tier, costs 20% less. Those cuts also change how usage counts against paid subscriptions in Codex and ChatGPT Work.
That is the business story. Flagship Sol still sits at the top for hard reasoning. Luna and Terra are the volume dials. If your firm was defaulting everything to the most expensive model "just in case," the math just got worse for that habit—and better for a deliberate model map.
OpenAI frames the cuts as the payoff from stack-level efficiency: models that take a more direct path through work, inference that squeezes more output from the same hardware, and an agentic harness that cuts repeated context and tool churn. Related OpenAI technical notes also claim Sol-assisted kernel work cut end-to-end serving cost by about 20% and raised token-generation efficiency by more than 15%. Treat those as vendor measurements until you verify them on your own workloads.
Cheaper tokens do not automatically mean a cheaper firm. Wrong-tier routing can erase the discount in one afternoon of retries.
Owner-led professional firms feel this first as invoice surprise. A paralegal workflow on Sol, a document triage job that Luna could finish, and a client summary that Terra would have handled—all billed as one "AI spend" line nobody can explain. OpenAI's own framing is that Luna can run tools and multi-step workflows at high volume, which is exactly where uncontrolled agents burn money if nobody names an owner.
OpenAI also points API customers to Fast mode on Sol when response time matters. Latency is another axis on the same ladder: not every job needs max reasoning, and not every job can wait. A firm that cannot tell those jobs apart will overpay for both speed and intelligence.
Price changes also reset continuity planning. Yesterday's "Terra is the bargain" spreadsheet is stale. So is a policy that says "use GPT" without naming Sol, Terra, or Luna. Model Selection is not a one-time purchase. It is a living shortlist with a fallback when a tier reprices, throttles, or disappears from your plan.
And if Codex or ChatGPT Work seats are in the mix, remember the cuts also change subscription burn rate. Seat count times wrong default model is how a "cheap" AI program becomes a quiet opex leak.
Treat the new Luna and Terra prices as a forcing function to map work to tiers—not as permission to turn everything on.
Starting today, GPT-5.6 Luna will cost 80% less, while GPT-5.6 Terra will cost 20% less. - OpenAI
This story maps to the two decisions most owner-led firms skip when a lab drops prices: which model for which job, and who keeps watching the meter.
Model Selection & Continuity Planning matches the model to the work—cost versus capability versus privacy—with a fallback so a reprice or restriction does not break the business. Sol, Terra, and Luna are a menu, not a mandate.
Managed AI Operations is the destination after the map is clear: monitoring, governance, and plain-English reporting so cheaper tiers do not quietly become unowned agent spend.
We stay vendor-neutral. The question is not whether OpenAI's cut is clever. It is whether your firm can explain which work runs on which tier—and what it returned.
If Luna and Terra just got cheaper in your stack and you cannot say which workflows should move, start with a short free assessment. No deck. Just clarity on cost, fit, and who owns the operating layer.
This article summarizes publicly reported information and is for general informational purposes only. It does not constitute legal, tax, financial, investment, security, or compliance advice. AgentsROI.ai is not a law firm, accounting firm, or registered investment adviser. Facts, pricing, statistics, and product capabilities cited here reflect the sources listed at the time of writing and may change. Readers should verify current information independently and consult qualified professionals regarding obligations specific to their industry, jurisdiction, and circumstances—including applicable New York State and New York City requirements. AgentsROI.ai may have commercial relationships with vendors mentioned; where material, such relationships are disclosed. Nothing in this article is an endorsement of any specific AI product, model, or provider.
