Chinese AI company DeepSeek has solidified its aggressive pricing strategy by making a massive 75% discount on its V4-Pro API permanent. Initially launched as a temporary campaign, the reduced rates will now remain indefinitely for developers worldwide. This move deepens the ongoing AI price war and pressures global competitors to slash their own API costs.
Under the new structure, input pricing for DeepSeek V4-Pro starts at $0.435 per million tokens, while output pricing is around $0.87 per million tokens—roughly a quarter of the model’s previous rates. According to third-party benchmark firm Artificial Analysis, the model ranks top globally for intelligence per dollar cost. DeepSeek’s offering is reportedly 10 to 35 times cheaper than leading US systems; running the Intelligence Index benchmark on V4-Pro costs about $268, compared to 12 times more for OpenAI’s GPT-5.5 and 19 times more for Anthropic’s Claude Opus 4.7.
The model supports one-million-token context windows, making it attractive for coding, mathematics, and agentic workloads. It also features OpenAI-compatible API integration, simplifying migration for developers. This cost advantage is reshaping enterprise AI adoption, with startups and independent developers increasingly prioritizing affordability over brand recognition.
The announcement comes as Chinese AI firms like DeepSeek, Qwen, Kimi, MiniMax and Xiaomi challenge US dominance through price and capability. While China’s broader economy remains soft—retail sales grew at their slowest pace since the post-COVID reopening—investors continue to pour money into the AI supply chain. Analysts note that many hardware and software companies in the ecosystem are profitable, though their scale is not yet large enough to lift the entire economy. The pricing war is expected to accelerate global enterprise AI deployment and give businesses more choices, while intensifying competition across the industry.
