Via blog.filestack.com
The Chinese AI lab that sparked a global price war is now hiking its own rates, with output token costs jumping more than tenfold on its premium model.
DeepSeek spent the first half of 2026 being the disruptor that made every other AI company sweat. Now it’s joining the club it once threatened to blow up.
The Chinese AI startup announced on August 6 that it will significantly raise API prices across its V4 model lineup, implementing a new peak and off-peak billing structure starting August 16. The most striking number: the V4-Pro model’s output token price during peak hours will climb to $3.96 per million tokens, up from the current $0.87. That’s roughly a 4.5x increase on the output side alone.
From price slasher to price matcher #
To understand why this matters, rewind a few months. In May 2026, DeepSeek made a permanent 75% price cut on its V4-Pro model, a move that sent shockwaves through China’s AI industry and rattled Western competitors.
The current rates tell the story of just how aggressively DeepSeek had been pricing. V4-Flash sits at $0.14 per million uncached input tokens and $0.28 for output tokens. V4-Pro runs $0.435 for input and $0.87 for output.
But cheap access to models that can handle contexts up to 1 million tokens creates a predictable problem. The new pricing structure, with its peak and off-peak tiers, suggests DeepSeek is now grappling with the reality that GPU hours cost real money regardless of how efficiently your architecture runs.
The ripple effects across China’s AI market #
DeepSeek didn’t just set its own prices low. It effectively set the floor for China’s entire AI API market. When the startup slashed rates, domestic competitors like Alibaba’s Qwen, Baidu’s Ernie, and a host of smaller players felt compelled to match or undercut those numbers.
For the thousands of Chinese developers and startups building on these APIs, the calculus changes materially. A V4-Pro output price of $3.96 per million tokens during peak hours is a fundamentally different economic proposition than $0.87. The V4-Flash model, which was already the budget option in DeepSeek’s lineup, may see increased adoption as developers look to manage costs. The exact new pricing for Flash hasn’t been disclosed yet.
What this signals about AI economics #
DeepSeek’s V4-Flash model launched its public beta just a week before the price hike announcement, on July 31. Releasing a new model and then immediately announcing higher prices is a classic sequencing move: get developers hooked on the capability first, then adjust the economics.
Developers building on DeepSeek’s APIs have ten days to prepare before the new billing structure kicks in on August 16. Those running cost-sensitive applications would be wise to audit their token usage patterns and figure out which workloads can be shifted to off-peak hours.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our