DeepSeek V4 launches mid-July with peak-hour API pricing surge

DeepSeek has officially announced that the full release of DeepSeek V4 will arrive in mid-July 2026, bringing a major shift in API pricing. The new model introduces double rates during peak hours — defined as 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM Beijing time. This surge pricing is designed to flatten demand spikes, improve server stability, and encourage developers to offload non-critical tasks to off-peak windows. The move signals a maturing infrastructure strategy from DeepSeek as it scales its user base globally.

Alongside the pricing overhaul, DeepSeek V4 will deliver performance optimizations for both the V4 Pro and V4 Flash variants. While exact benchmarks and new feature details remain under wraps, the company promises efficiency gains and lower latency for routine operations. For developers, the key trade-off becomes clear: pay a premium for guaranteed peak-time throughput, or restructure workflows to benefit from cheaper off-peak access. This tiered pricing is likely to segment heavy and light users, potentially reshaping how AI workloads are scheduled across the ecosystem.