It's as if the AI industry coordinated a script in early July.
On July 8, Alibaba Cloud suddenly announced time-based pricing for Qwen3 series—peak-hour API calls up 30%, off-peak down 10%. In plain terms: daytime usage, expensive; late-night usage, cheap.
Same day, DeepSeek V4's official pricing took effect. Output tokens jumped from ¥2 to ¥4 per million—doubled overnight. Developer communities erupted in frustration: "Was paying ¥300 a month, now it's ¥600. Didn't do anything different."
July 9, GPT-5.6 unlocked globally. Sol/Terra/Luna—three models launched simultaneously, stunning performance. But OpenAI also announced the Pro subscription rising from $200/month to $300/month, with API pricing adjusted upward in tandem.
Three headlines, one picture: large models are hiking prices across the board. And this is just the beginning.
{
It's Not About One Company—the Whole Industry Is Closing the Net
Look at the trajectory.
2023-2024 was AI's "land-grab era." ChatGPT was free, Claude was free to try, every company's API pricing was a subsidy-fueled price war. Everyone was competing for users, nobody was thinking about profit. Using one model, two models, ten models—cost was negligible.
2025 was the inflection point. DeepSeek announced commercialization in early January, Claude rolled out tiered subscriptions, Gemini tightened free quotas. By late June and early July, collective price hikes became an open secret.
Why? Because the AI industry can't keep burning money. Training a single large model costs tens of millions of dollars, and inference costs remain stubbornly high. VC money isn't charity—someone has to pay eventually.
And who's that someone?
It's you. It's your monthly API bill. It's that page where the deduction number keeps getting bigger.
{
Prices Are Rising—Most People Will Just Take It. You Don't Have To.
What makes price hikes scary isn't "paying more." It's being locked in. You've been using DeepSeek, your APIs are integrated, your workflows are built, your Agent runs on it. They double the price—you grit your teeth and pay. Because switching costs are too high.
But what if your AI could switch models anytime?
Kaihe AIBOX is a 24/7 local Agent hardware box. It isn't bound to any single model. GPT-5.6 got more expensive? Switch to an open-source model. DeepSeek doubled prices? Run Qwen3 on Kaihe first, then deploy open-source M3 Pro locally when it drops—zero API fees. Alibaba's peak pricing hitting you? Schedule your Agent to run at 3 AM during off-peak hours—wake up to results at a fraction of the cost.
Models can raise prices all they want. You can switch all you want.
This isn't about saving money. It's about taking pricing power back from the model companies.
AI is your infrastructure. You shouldn't be just another line item on some company's monthly billing sheet.
Want to learn more about Kaihe AIBOX? Contact us at [email protected]
KaiheAIBOX #AIPriceHike #GPT5.6 #DeepSeek #AlibabaCloud #AICosts #LocalAgent
📖 Glossary
AI Box (also known as Agent Computer / Agent PC), is a dedicated local hardware device that runs AI Agents. Pre-installed with an AI agent management system, plug-and-play, running 24/7. Users can remotely command AI to work via Discord, Slack, Telegram, WhatsApp, and more.