OpenAI has sharply lowered prices across its GPT-5.6 frontier model lineup, cutting GPT-5.6 Luna by 80 percent and GPT-5.6 Terra by 20 percent while introducing a premium Fast mode for the flagship Sol model. The move places Luna at $0.20 per million input tokens and $1.20 per million output tokens, for a combined $1.40 per million tokens.
These reductions arrive days after Anthropic launched Claude Opus 5 at the same price as its predecessor and Google released Gemini 3.6 Flash and Gemini 3.5 Flash-Lite with emphasis on lower inference costs. OpenAI now undercuts Google on price per unit of intelligence while offering speed advantages that may appeal to Anthropic customers.
Luna’s new pricing moves the smallest GPT-5.6 model into direct competition with the market’s lowest-cost commercial offerings. Although Xiaomi’s MiMo-V2.5 Flash and DeepSeek’s flash variant remain cheaper on raw token rates, Luna now sits below Google’s Gemini 3.5 Flash-Lite at $2.80 combined and far below Gemini 3.6 Flash at $9.
The adjustment widens the gap between OpenAI’s three tiers. Luna now costs one-tenth the combined price of Terra, which itself dropped to $14 per million tokens to match Google’s Gemini 3.1 Pro Preview for contexts up to 200,000 tokens. Sol Standard stays at $35 combined, while the new Fast mode doubles that rate to $70 for up to 2.5 times higher throughput.
One key implication is that enterprises running high-volume workloads such as document summarization and real-time classification can now achieve frontier-level performance at dramatically lower operating expense. This shift lowers the barrier for deploying AI agents in cost-sensitive sectors where token volume previously made large-scale use uneconomical.
Another consequence is increased pressure on total cost of ownership calculations. Providers are no longer competing solely on intelligence benchmarks; they are optimizing the economics of completing entire production tasks. OpenAI’s direct per-token cuts, paired with Google’s focus on reduced token usage and Anthropic’s capability gains at unchanged rates, give buyers multiple levers to control expenses.
The GPT-5.6 series was initially released in limited form in late June 2026, with Sol aimed at complex reasoning, Terra at balanced production use, and Luna at high-throughput, low-latency jobs. The latest pricing changes reposition Luna as OpenAI’s explicit answer to the growing low-cost segment without requiring a new model generation.
By moving Luna into the low-cost tier while preserving Sol’s premium positioning, OpenAI signals that frontier capability is now available across a broader range of budgets. The changes underscore how quickly model economics, rather than mere access, have become the central battleground among leading AI providers.
