Beijing, September 9, 2026 — Chinese AI company DeepSeek announced on September 9 that it will cut API prices for its Flash-series models by up to 60 percent, effective September 10. The deepest reduction targets cache-hit input pricing, which drops to RMB 0.02 per million tokens (about USD 0.003).
The move comes less than four weeks after DeepSeek raised rates on its V4 Pro line by as much as 11-fold. Output pricing, however, remains at double the pre-August level, preserving a margin buffer on the generation side.
Concurrent with the announcement, DeepSeek opened a 48-hour closed beta for V4.1 Flash — a structurally redesigned model with native multimodal support — distributed via developer community channels with a concurrency cap of 20 sessions per account. Early developer feedback flagged inference speeds up to six times faster than the current Flash Vision model.
Separately, Reuters reported that DeepSeek has retained CITIC Securities to underwrite a planned IPO on Shanghai's STAR Market, with a private round targeting a pre-money valuation of RMB 500 billion (about USD 69-71 billion). The company generated roughly RMB 475 million (USD 65.9 million) in revenue in the first seven months of 2026, implying a revenue multiple exceeding 140x at the upper end of the valuation range.
