The AI Pricing Landmine: DeepSeek Jacks Prices 93% While GPT-5.6

DeepSeek raised its V4 Flash API price by 93 percent on August 14, from $0.14 to $0.27 per million input tokens, while OpenAI cut GPT-5.6 Luna's price by 80 percent to $0.20 per million and made it the free default for ChatGPT on August 15. The moves signal the end of stable AI pricing, with developers warned to treat launch prices as promotional.

The AI Pricing Landmine: DeepSeek Jacks Prices 93% While GPT-5.6 The AI pricing landscape shifted faster in the last two weeks than at any point since frontier models launched: DeepSeek raised V4 Flash 93% to $0.27 per… Two price moves in the space of a single day have quietly reshaped the economics of building with AI, and developers who missed them are about to get a bill they didn't expect. On August 14, DeepSeek /compare/llama-4-vs-deepseek-r1 raised its V4 Flash API price by 93 percent, from $0.14 to $0.27 per million input tokens. A day later, OpenAI /glossary/openai made GPT /glossary/gpt -5.6 Luna the free default for ChatGPT /compare/chatgpt-vs-claude , after cutting its price 80 percent at the end of July from $1 to $0.20 per million. Opposite directions, same market. The AI pricing landscape has moved faster in the last two weeks than in any comparable period since frontier models first launched. DeepSeek's 93 Percent Hike DeepSeek's V4 Flash was released in late July as the company's agent-coding workhorse, priced aggressively at $0.14 per million input tokens and $0.28 for output. The August 14 revision lifts input pricing 93 percent to $0.27, with output pricing moving in parallel. The timing matters. DeepSeek undercut the market to build developer adoption of the official V4 Flash build 0731, then repriced once that adoption took hold. It's a classic land-and-expand move, executed within a three-week window. DeepSeek hasn't framed the increase as anything other than a pricing update, but the message is clear: the introductory rate was exactly that. Developers who built their cost models on the July numbers are now paying close to double. OpenAI Going the Other Way OpenAI's move is the mirror image. GPT-5.6 Luna became the free ChatGPT default on August 15, with unlimited chats, following an 80 percent price cut on July 30 that brought it from $1 to $0.20 per million tokens. OpenAI is effectively subsidizing its own free tier to defend consumer share against the wave of cheaper, open- weight /glossary/weight alternatives that shipped this summer. The asymmetry is the point. OpenAI can afford to give away Luna for free because its paid tiers - GPT-5.6 Sol Max and the enterprise products - carry the margin. DeepSeek has no comparable high-end tier to absorb losses, so its cheaper model has to start paying for itself. What This Means for Builders For developers, the lesson is blunt: don't build your cost model on an API's launch-day price. The gap between introductory pricing and sustainable pricing has collapsed from quarters to weeks. Three practical implications follow. First, treat every new model's launch price as promotional until proven otherwise. Second, build abstraction layers so a 93 percent price hike doesn't force a rewrite. Third, watch the free tiers carefully - OpenAI giving away Luna reshapes the floor for what consumers expect to pay, which ripples up into the rates freelance knowledge workers can charge for work that free AI can now approximate. The era of stable AI pricing is over before it really began. The only safe assumption is that whatever you're paying today won't hold a month from now. Sources: AI Tools Recap daily briefings, August 14-15, 2026; DeepSeek pricing changelog, August 14, 2026; OpenAI product update, August 15, 2026. Get AI news in your inbox Daily digest of what matters in AI.