OpenAI Cuts GPT-5.6 Luna Prices by 80% as AI Competition Shifts Toward Cost

AI News

July 31, 2026

OpenAI Cuts GPT-5.6 Luna Prices by 80% as AI Competition Shifts Toward Cost

OpenAI has reduced the price of GPT-5.6 Luna by 80%, intensifying competition among artificial intelligence providers seeking to make advanced models affordable for large-scale business applications.

The company also lowered GPT-5.6 Terra prices by 20% and introduced a faster processing option for its flagship GPT-5.6 Sol model. The changes took effect on July 30, only weeks after the GPT-5.6 family became broadly available.

Luna becomes OpenAI’s cheapest GPT-5.6 model

GPT-5.6 Luna is designed for cost-sensitive workloads that require fast responses and large volumes of requests.

The standard API price has fallen from $1 to $0.20 per million input tokens. Output pricing has dropped from $6 to $1.20 per million tokens for requests using the standard short-context tier.

For long-context requests, Luna now costs $0.40 per million input tokens and $1.80 per million output tokens. Cached input receives an additional discount, reducing the price to $0.02 per million tokens for standard short-context processing.

The reduction makes Luna slightly cheaper on output than GPT-5.4 nano, which costs $1.25 per million output tokens, while offering the newer GPT-5.6 architecture and support for agentic workflows.

Terra prices also fall

OpenAI has cut the standard price of GPT-5.6 Terra from $2.50 to $2 per million input tokens and from $15 to $12 per million output tokens.

Terra sits between Luna and the more powerful GPT-5.6 Sol. It is intended for everyday professional work where users need stronger reasoning than Luna provides but do not require the maximum capabilities of the flagship model.

OpenAI says the lower prices will also affect how Luna and Terra usage is counted within paid Codex and ChatGPT Work subscriptions. This could allow subscribers to complete more tasks without reaching their usage limits as quickly.

OpenAI targets high-volume AI agents

The price cuts are particularly important for companies operating AI agents.

Unlike a simple chatbot response, an agent may make repeated model calls while searching documents, using software tools, checking its work and completing multiple steps. Even small differences in token pricing can therefore produce substantial changes in operating costs.

OpenAI says Luna can use tools and complete multi-step workflows, making it suitable for background automations, routine coding, data processing and other high-volume tasks.

Several early customers report using Luna as a lower-cost component alongside more capable models. A workflow might use GPT-5.6 Sol to plan a complex task and delegate implementation, testing or evaluation to Luna.

This model-routing approach allows businesses to reserve expensive frontier intelligence for the parts of a workflow where it creates the most value.

Sol receives a faster API mode

OpenAI has not reduced the base price of GPT-5.6 Sol.

Instead, the company introduced Fast mode, which provides up to 2.5 times faster processing at twice the standard API price. The option replaces Priority Processing, although existing requests using the previous priority setting will continue to work.

Sol remains priced at $5 per million input tokens and $30 per million output tokens under standard short-context processing. Fast mode increases those rates to $10 and $60 respectively.

The option is intended for applications where response time is more important than minimizing inference costs.

Efficiency improvements drive lower prices

OpenAI attributes the reductions to improvements across model architecture, inference infrastructure and context management.

The company says GPT-5.6 Sol helped optimize production software and model-serving kernels. That work reportedly reduced end-to-end serving costs by 20%, while separate experiments improved token-generation efficiency by more than 15%.

Better context management can also reduce repeated processing in long-running agent workflows. When models reuse cached instructions and previously supplied information, businesses pay less than they would for repeatedly processing the same full prompt.

OpenAI is passing part of those infrastructure savings to customers through lower API prices.

AI providers compete on cost

The announcement highlights a broader change in the AI market.

Model providers initially competed mainly through benchmark results, context-window sizes and reasoning capabilities. As performance differences narrow, API price, latency and total cost per completed task are becoming more important to enterprise customers.

Google currently charges $1.50 per million input tokens and $9 per million output tokens for Gemini 3.5 Flash. Anthropic launched Claude Sonnet 5 at an introductory price of $2 per million input tokens and $10 per million output tokens.

These models differ in performance, capabilities and intended use, so token prices cannot be compared in isolation. A cheaper model may become more expensive if it needs additional prompts, retries or human review to produce an acceptable result.

OpenAI itself argues that companies should measure cost per successful outcome rather than focusing only on the price of individual tokens.

Cost becomes a competitive advantage

The GPT-5.6 Luna price cut gives OpenAI a stronger option for businesses processing millions or billions of tokens through automated systems.

It may also put pressure on competitors to adjust prices, introduce smaller models or offer additional discounts for batch processing and cached prompts.

The next phase of AI competition is unlikely to be decided by intelligence alone. For many enterprise applications, the winning model may be the one that delivers sufficient quality at the lowest total cost, highest speed and greatest operational reliability.

ChatGPT OpenAI