OpenAI’s price shock: Luna at $0.20 per million tokens
OpenAI’s new GPT-5.6 Luna and Terra pricing is a major API cost reduction that slashes token rates, reshapes how developers plan workloads, and forces enterprises to rethink long-term AI budgets around cheaper, high-volume automation. OpenAI has cut GPT-5.6 Luna API prices by 80%, dropping input tokens from USD 1 (approx. RM4.60) to USD 0.20 (approx. RM0.92) per million and output tokens from USD 6 (approx. RM27.60) to USD 1.20 (approx. RM5.52) per million. That is not a tweak; it is a reset of what top-tier models cost at scale. Terra gets a smaller but still meaningful 20% cut, from USD 2.50 (approx. RM11.50) to USD 2 (approx. RM9.20) for input and from USD 15 (approx. RM69.00) to USD 12 (approx. RM55.20) for output per million tokens. In practical terms, workloads that were marginal now become affordable experiments.
Why OpenAI can afford these cuts: efficiency as a weapon
These OpenAI price cuts are not charity; they are the direct output of technical efficiency and a growth-first strategy. OpenAI says the reductions were enabled by efficiency improvements across its models and serving systems. Within a human-led process, GPT-5.6 Sol autonomously rewrote and optimized production GPU kernels, cutting end-to-end serving costs by 20% and improving speculative decoding so token generation became more than 15% more efficient. According to OpenAI, “As we get more efficient, we are passing the savings on to you!”. That line is both marketing and a warning shot: they are ready to compete on price as much as on model quality. With Sol’s new Fast mode delivering up to 2.5x speed at twice the standard cost while keeping intelligence unchanged, OpenAI is openly turning latency and price into configurable dials rather than fixed constraints.
One billion users and the new AI adoption flywheel
The timing is not accidental: OpenAI announced these GPT-5.6 Luna and Terra cuts one day before revealing that its models now reach more than one billion active users and more than two million businesses. The company says engagement grows over time, with people sending around 50% more messages per day and using ChatGPT for roughly twice as many kinds of work after six months. In other words, AI use deepens once it enters a workflow—so lowering the entry cost is a direct bet on volume. “Better intelligence drives broader adoption. Broader adoption supports more investment. More investment improves intelligence and efficiency,” wrote OpenAI CFO Sarah Friar. This is the classic flywheel story. Yet the risk is obvious: OpenAI generated USD 13.07 billion (approx. RM60.12 billion) in 2025 revenue but lost USD 21 billion (approx. RM96.60 billion). Cheaper APIs may speed adoption while delaying profitability.
What developers gain: cheaper experiments, new architecture choices
For developers, GPT-5.6 Luna pricing at USD 0.20 (approx. RM0.92) per million input tokens and USD 1.20 (approx. RM5.52) per million output tokens is transformative for high-volume work. Luna is pitched as the fastest and lowest-cost GPT-5.6 model that can use tools and complete multi-step workflows. That means workloads that once needed careful rationing—log analysis, synthetic data, bulk content generation—can move from “premium feature” to everyday default. Luna and Terra now also consume fewer credits in ChatGPT Work and Codex, while subscription prices and quotas remain unchanged. This effectively stretches existing budgets without any contract renegotiation. Architecturally, the new landscape encourages tiered design: use Luna for routine or high-volume processing, Terra as the middle tier, and Sol (with optional Fast mode) only when latency or absolute peak capability matters. The result is a more nuanced cost-performance stack than the old one-model-for-everything approach.
Enterprise AI budgets: more room to build, less room for excuses
For enterprises, these API cost reductions remove one of the last defensible reasons to stay on the sidelines. Luna and Terra becoming less expensive for routine or high-volume processing while Sol offers a paid speed boost means buyers now have a clear and transparent trade-off between speed, capability and price. On top of that, the same Luna and Terra price cuts flow into ChatGPT Work and Codex credit accounting, lowering model consumption without raising subscription fees. Lower prices could encourage even wider adoption, though analysts warn they could put more pressure on margins as OpenAI keeps spending heavily on computing infrastructure. For enterprise AI budgets, this is both an opportunity and a signal. The opportunity: shift savings into broader deployments and more ambitious use cases. The signal: AI platforms are in a price war, and locking into a single vendor without revisiting costs regularly now looks like financial negligence.
