DeepSeek Permanently Slashes V4 Pro AI Model Price by 75%, Escalating AI Price War

DeepSeek, a prominent Chinese artificial intelligence startup, has announced a permanent 75% reduction in the price of its flagship V4 Pro AI model. This significant price cut, which was initially a promotional offer set to expire at the end of May 2026, has now been made a permanent fixture, dramatically altering the competitive landscape of the AI model market. Under the new pricing structure, DeepSeek V4 Pro's costs range from approximately $0.003625 to $0.87 per million tokens, a substantial decrease from its previous range of $0.0145 to $3.48 per million tokens. This makes DeepSeek V4 Pro considerably more affordable than its Western counterparts. For instance, OpenAI's GPT-5 charges $2.50 per million input tokens and $10 per million output tokens, while Anthropic's Claude Opus 4.7 is priced at $5 for input and $25 for output. Even Google's cost-optimized Gemini 3.5 Flash, at $0.15 input and $0.60 output per million tokens, is undercut by DeepSeek's new rates. The decision to lock in such a deep discount, just a month after the V4 models were launched, suggests a clear strategy by DeepSeek to prioritize market share over immediate per-unit revenue. This aggressive pricing is expected to appeal to enterprise customers and power users who process millions of tokens daily, offering substantial cost savings. The company had previously indicated that pricing for the Pro version would decrease once Huawei Ascend 950 supernodes became widely available in the latter half of the year, suggesting that improved access to high-end computing capacity might be a factor behind this permanent price reduction. However, the competitive dynamics are further complicated by an unresolved accusation from Anthropic, which alleged that DeepSeek engaged in “distillation attacks,” improperly training its models on Claude's responses. If substantiated, this would imply that some of DeepSeek's performance advantage could stem from leveraging Anthropic's research investments, making the price differential a matter of intellectual property arbitrage rather than pure engineering efficiency. DeepSeek's move intensifies the ongoing AI price war, pushing the broader AI pricing landscape further towards commoditization throughout 2026. This strategy aims to attract significant volume to the cheapest capable model, regardless of its origin, potentially bifurcating the market into Western and Chinese tiers with distinct economic models.