What DeepSeek’s Permanent 75% Price Cut Actually Means
DeepSeek’s permanent 75% price cut for its V4-Pro model is a long-term reduction that fixes API pricing at one-quarter of the original launch level, signaling a structural shift rather than a temporary promotion in how advanced AI capabilities are priced and accessed by developers and businesses. According to Technology.org, “DeepSeek’s V4-Pro now costs a quarter of its launch price, and the company says that rate is here to stay.” The new AI model pricing range for V4-Pro is 0.025 to 6 yuan per million tokens, far below the earlier 0.1 to 24 yuan band. In practical terms, the affordable AI API now spans roughly $0.0035 to $0.83 per million tokens. For developers watching every cent of inference cost, this V4-Pro cost reduction moves a flagship model into a price tier that used to belong only to lighter, stripped-down systems.
Challenging Premium AI Model Pricing and Business Models
By locking in this 75% DeepSeek price cut instead of treating it as a launch incentive, the company is directly challenging the idea that top-tier models must carry premium, volatile price tags. At launch, V4-Pro had been positioned as a significantly more expensive option, with DeepSeek warning that the Pro tier could cost up to 12 times more than the lighter Flash version because of “constraints in high-end compute capacity.” Now, pricing parity is much closer, and that change forces competitors to justify higher per-token rates on performance, tooling, or ecosystem alone. This move also changes how teams budget for AI. When an affordable AI API is not a limited-time discount but a permanent price floor, long-term product planning, cost forecasting, and multi-year contracts become less risky for engineering leaders.
Scaling, Huawei Ascend 950, and DeepSeek’s Market Signal
The decision to make the V4-Pro cost reduction permanent also sends a strong signal about DeepSeek’s confidence in its scaling path and hardware supply. The model relies on Huawei’s Ascend 950 chips, which DeepSeek had earlier linked to future price drops once those chips shipped in higher volume in the second half of the year. While the company did not explicitly confirm that a steadier flow of Ascend 950 hardware triggered this change, the timing suggests improved access to compute and better efficiency. US export rules blocking Nvidia’s most advanced chips have steered local demand toward Huawei’s AI silicon, even as separate restrictions slow Ascend production growth. In that context, DeepSeek’s permanent pricing looks like a strategic bet that its infrastructure will stay efficient enough to support lower AI model pricing while still competing with more established platforms.
How Lower Inference Costs Expand AI Adoption
For startups, small businesses, and cost-sensitive applications, the new V4-Pro pricing tier changes what is financially realistic. Inference-heavy products such as chat interfaces, code assistants, or batch document processing tools often see token usage measured in tens or hundreds of millions; shifting those workloads to an affordable AI API that charges 0.025 to 6 yuan per million tokens can materially change unit economics. Lower, predictable AI model pricing makes it easier to experiment with new features, support more generous free tiers, or run internal tools without pushing cloud bills over budget. It also narrows the trade-off between model power and cost, so teams no longer have to default to weaker models when scaling user numbers. Over time, this could shift the competitive landscape, as smaller players gain access to capabilities that once required deep pockets.