Skip to main content
GPUBeat Frontier Models DeepSeek Enters Price War with Permanent…

DeepSeek Enters Price War with Permanent 75% Discount on V4-Pro

DeepSeek has made a bold move by permanently slashing prices on its V4-Pro model, intensifying competition with major players like OpenAI and Anthropic.

DeepSeek has made a notable change in its pricing strategy, announcing that the 75% discount on its V4-Pro API output tokens will remain in effect indefinitely. This decision, revealed on Saturday, positions the company's offering as a strong competitor to established players in the AI infrastructure market, such as OpenAI and Anthropic.

The new pricing structure for the V4-Pro model drops the cost to $0.87 per million tokens, a significant decrease from the launch price of $3.48 just one month ago. This aggressive pricing places DeepSeek in a distinctive position, undercutting the competition by a substantial margin. For example, OpenAI's GPT-5.5 charges $30 per million tokens, while Anthropic's Claude Opus 4.7 costs $25 for the same output.

Market Impact

The implications of this price cut are far-reaching. With V4-Pro now priced as low as $0.0035 per million tokens depending on usage, applications generating substantial output can see considerable savings. For instance, a typical application producing 100 million output tokens per month would incur costs of roughly $87 on V4-Pro, compared to $2,500 on Claude Opus 4.7 or $3,000 on GPT-5.5.

This pricing strategy not only highlights DeepSeek's ambition to capture market share but also indicates a shift in the space of AI services. As other players respond, the cost disparity could drive adoption of DeepSeek's AI solutions among developers and businesses seeking economical alternatives.

Competitive Landscape

The AI market is rapidly changing, and DeepSeek's decision comes at a time of increasing competition. According to CloudZero, Google's Gemini 2.5 Pro is priced at $12 per million tokens, while Claude Sonnet and GPT-4.1 are priced at $15 and $8, respectively. This effectively positions DeepSeek as a low-cost leader in the space, potentially attracting a wider range of customers.

The permanence of this discount arrives ahead of the anticipated boost in high-end computing capacity expected with the release of Huawei Ascend 950 supernodes, projected for the latter half of 2026. While DeepSeek did not specify whether these chips contributed to the price cut, the decision to maintain low rates before hardware advancements is noteworthy.

Future Outlook

As AI infrastructure continues to develop, the sustainability of DeepSeek's pricing strategy will be essential. The company is not only challenging existing pricing models but also establishing a new benchmark for cost efficiency in AI services. This could prompt further price adjustments across the industry as competitors react to DeepSeek's bold approach.

In light of these developments, the AI market is poised for significant changes. The lower pricing could encourage innovation and drive new applications, as crypto bots and DeFi agents finally find pricing that suits their needs. As the market observes closely, DeepSeek's upcoming moves will likely influence the future of AI infrastructure pricing.

See also  Ex-Google DeepMind Researcher Highlights AI Evaluation Crisis
GD

GPUBeat Desk

Desk · joined 2026

GPUBeat Desk covers AI infrastructure — chips, foundation models, inference economics, datacenter buildouts, and the geopolitics of compute.