AI Pulse · Frontier Labs Watch

DeepSeek just changed how it charges for its models, and the change is bigger than a routine price update. Starting August 16, the company is moving away from flat pricing and into peak and off-peak tiers. V4-Flash output jumps from a flat $0.28 per million tokens to $1.32 at peak, with a $0.66 off-peak rate. V4-Pro goes from $0.87 to $3.96 at peak, with $1.98 off-peak. DeepSeek says this is about capacity strain, meaning demand for their models has outpaced what their infrastructure can handle at the old prices. Details in the original report.

Here’s why this matters if you’ve built any cost models around AI. A lot of businesses, especially ones running high volume tasks like content generation, customer support, or data processing, picked Chinese models specifically because they were cheap and assumed to stay cheap. That assumption just took a hit. A more than tenfold jump at peak hours means your per-task cost could look very different depending on when your workload actually runs. If you’re running batch jobs at 2pm when everyone else is too, you’re paying premium rates for the privilege.

The practical move is to check whether your usage is peak-heavy or if you can shift workloads to off-peak windows and cut costs close to half. It’s also a reminder not to lock your architecture into one provider’s pricing structure, because that structure can change with two weeks notice. This is exactly the kind of shift that’s easy to miss until your bill shows up wrong, which is why we build usage and cost tracking into an AI command centre so you catch pricing changes before they catch you.