OpenAI announced an 80% price reduction for its GPT-5.6 Luna model, lowering API costs to $0.20 per million input tokens and $1.20 per million output tokens as of July 30, 2026 [1, 2, 3, 4, 5, 6, 7, 8, 9]. The company also cut prices for the mid-tier GPT-5.6 Terra model by 20%, setting fees at $2.00 per million input tokens and $12.00 per million output tokens [1, 2, 3, 4, 5, 6, 7, 8, 9].
For the flagship GPT-5.6 Sol model, OpenAI did not reduce base prices but introduced a new Fast mode that boosts processing speed up to 2.5 times the standard rate at twice the cost, with no change to model capabilities [1, 2, 4, 5, 9]. OpenAI CEO Sam Altman said, "Major price cuts today. We want to offer the best price/intelligence tradeoff at every level" [1, 4].
The GPT-5.6 series, which includes Luna, Terra, and Sol models, was launched about three weeks earlier around July 9, 2026 [1, 3, 4, 6, 7]. The price cuts were made possible by infrastructure upgrades such as rewritten GPU kernels, improved inference efficiency, and better runtime optimizations that lowered serving costs by roughly 20% or more [4, 5].
OpenAI faces growing cost pressure from enterprise customers and increased competition from domestic rivals like Anthropic and Chinese firms such as Moonshot AI, whose Kimi K3 open-weight model is garnering attention [1, 3, 6, 7, 8]. The Luna model now ranks as the top intelligence-per-dollar option, surpassing competitors like Zhipu AI’s GLM-5.2 and MiniMax’s M3 [1].
The price reductions also impact usage-based pricing for OpenAI’s Codex and ChatGPT Work subscriptions [2, 5]. Despite the lower costs for Luna and Terra, Sol’s standard pricing remains unchanged, with its Fast mode providing a new premium speed option [1, 3, 4, 6, 7, 8, 9].
