OpenAI cuts GPT-5.6 Luna price by 80% as Chinese AI competition squeezes margins

OpenAI has cut prices on two of its GPT-5.6 models, effective July 30, just three weeks after the family launched on July 9. The fastest and most affordable model in the lineup — GPT-5.6 Luna — dropped 80%, from $1 per million input tokens to $0.20, and from $6 to $1.20 per million output tokens. The mid-range GPT-5.6 Terra saw a 20% reduction, settling at $2 per million input and $12 per million output. The flagship GPT-5.6 Sol remains unchanged.

Speed option added for Sol

Alongside the price cuts, OpenAI introduced an optional Fast mode for Sol: users can get 2.5x faster output at 2x the standard Sol cost. ChatGPT and Codex subscription prices and quota budgets are unchanged. For API users on Terra and Luna, usage now consumes fewer API credits per call, in line with the new pricing.

Why the speed of the reversal is notable

A major price cut three weeks after launch is unusual even by AI industry standards, where pricing has been falling steadily since GPT-4 launched in 2023. It reflects the degree to which Chinese AI models — particularly DeepSeek, Qwen, and MiniMax — have compressed the premium that US frontier labs can charge for raw inference performance. Chinese models now routinely match or exceed Western benchmarks at a fraction of the cost, and that competition has forced US providers to recalibrate faster than they anticipated when GPT-5.6 launched.

The Luna cut is especially telling. Luna was already positioned as the low-cost tier of the GPT-5.6 family; cutting it another 80% within three weeks suggests OpenAI saw significant developer churn to cheaper alternatives at the old price point. At $0.20 per million input tokens, Luna is now priced to compete directly with the cheapest models from Anthropic, Google, and the Chinese providers on a cost-per-call basis.

Competitive context

The price move comes as the broader AI model market is undergoing structural compression. Amazon Web Services posted 37% cloud growth in Q2 and raised its 2026 infrastructure spend to $220 billion, but even with that demand surge, model pricing continues to fall because the supply of capable inference capacity has expanded faster than premium demand. The price war is no longer just between labs — it now runs through the hyperscalers, Chinese providers, and open-weight alternatives simultaneously.

OpenAI has separately launched a free academic research program offering GPT-5.6 Sol Pro access to 100,000 scientists and engineers — a move that expands its research footprint while keeping commercial pricing pressure contained to a separate tier.

Price details confirmed via OpenAI's pricing announcement, with market context from CNBC.