DeepSeek raised V4 Flash pricing by 93%
The specific figure behind the trend reported yesterday: a 93% increase on V4 Flash, effective 14 August, in the same week OpenAI made a current model free without limits and Google discounted Flash by half.
Yesterday the observation was that DeepSeek was moving against the market. The number is now available and it is not a marginal adjustment — a 93% increase is close to a doubling, on the model tier that carries the highest volume.
A near-double on a high-volume tier is a repricing of the customer base, not a tweak. It will move workloads, and DeepSeek will learn from the migration exactly how much of its adoption was capability preference and how much was price.
That is the experiment, and it is a reasonable one to run. A lab that has shipped repeatedly, published weights and built genuine developer trust has options it did not have eighteen months ago, and being permanently cheapest is a position anyone with more capital can take away.
The contrast inside one week is now stark: 93% up here, a current model free with no chat limit there, and a 50% cut with a January doubling attached elsewhere. Inference pricing in 2026 has no trend line. It has four labs in four competitive positions.
AI Weekly — AI News Today, August 15 — Top AI Stories & Live Updates → · LLM Stats — LLM News Today (August 2026) →