Price Reset

Chinese artificial intelligence startup DeepSeek has made a permanent 75% reduction to the application programming interface price of its flagship DeepSeek-V4-Pro model.

On May 22, the company said a temporary developer promotion would become its regular pricing tier.

Under the new standard, the cache-miss input price for V4-Pro is 3 yuan per million tokens, or about $0.44.

Using the required currency format, that is about AED 1.53 for 3 yuan and about AED 1.61 for $0.44.

The new rate is far below the roughly $5 price attached to OpenAI’s GPT-5.5, or about AED 18.35, and also below the $0.95 charged by Chinese rival Kimi, or about AED 3.49.

The move sharpens price competition across both Chinese and overseas AI providers.

Based on the figures provided, DeepSeek’s new V4-Pro rate is less than one-tenth of the OpenAI price cited in the same comparison and also well under Kimi’s level.

Model Position

DeepSeek-V4-Pro is described as the world’s largest open-weight AI model with 1.6 trillion parameters.

At the same time, third-party benchmark provider VALS AI places the model ninth globally, with an average accuracy score of 63.87%.

Those details present a mixed picture: DeepSeek is pushing a very large model while also competing aggressively on API cost.

The benchmark standing included in the available material shows that the company is not framing the model only around size, but also around how it performs against other systems tracked by outside evaluators.

Additional usage data in the available material points to active demand on OpenRouter.

A source summary says DeepSeek’s market share there rose to 23.1%.

Separately, DeepSeek-V4-Flash ranked first on OpenRouter with 3.43 trillion tokens consumed.

Those figures refer to platform usage rather than the pricing change itself, but they add context to the company’s competitive position during the same period.

Funding Backdrop

DeepSeek is also pursuing its first round of external financing, with the valuation reported at $44 billion, or about AED 161.48 billion.

That fundraising backdrop gives the pricing change added significance because the company is adjusting commercial terms for its flagship model while also moving toward outside capital.

The available material does not provide further financing terms, investor names, or timing beyond the description that fundraising is looming.

Still, the combination of a permanent API price reduction and a first outside fundraising effort places commercial strategy and capital planning side by side.

Taken together, the reported developments are straightforward.

DeepSeek has locked in a steep discount for developers, bringing V4-Pro pricing down to 3 yuan per million tokens.

That undercuts the comparison prices listed for OpenAI’s GPT-5.5 and Kimi.

At the same time, the company is tied to a reported $44 billion valuation in an upcoming external funding round.

Alongside that, V4-Pro is presented as a 1.6 trillion-parameter open-weight model that ranks ninth globally on VALS AI with 63.87% accuracy, while DeepSeek’s broader model lineup shows strong OpenRouter usage metrics.