What happened with DeepSeek’s permanent 75% discount?

DeepSeek has made a 75% discount on its flagship AI model permanent, extending a price move that was originally set to expire.

The pricing change targets the V4 Pro model. In coverage of the update, DeepSeek said it will lower V4 Pro API prices by 75%, reducing costs to $0.435 per 1M input tokens and $0.87 per 1M output tokens. The discount was previously scheduled to end on May 31, but the company now plans to keep those reduced rates.

The announcement matters because it intensifies the ongoing AI model price war. Lower per-token costs can attract new developers and increase usage among existing customers, especially for applications that do lots of generation work—like customer support, coding assistants, and agentic workflows.

It also affects competitive dynamics: when one major provider makes pricing more aggressive and commits to keeping it, rivals typically face pressure to respond either with matching discounts, better performance-per-dollar, or alternative packaging (enterprise deals, different model tiers, caching features, or usage limits).

Separately, broader coverage in the same stream also points to how AI infrastructure cost pressures are reshaping the market: memory and compute constraints can push providers to change how they price or what they offer. In that context, DeepSeek’s move is notable because it signals confidence that the lower rates can persist even as cost and demand fluctuate.

Net impact: DeepSeek is locking in lower costs for V4 Pro API access, making it cheaper to build and scale AI apps and potentially forcing competitors to reevaluate their pricing.