Newsletters
Chinese artificial intelligence company DeepSeek released a new model with a significant price cut as competition among domestic technology firms intensifies.
Early benchmark results indicate V4-Flash could be among the least expensive artificial intelligence models to operate at scale globally.
According to Semafor, DeepSeek released the V4-Flash model as part of a strategy to regain momentum in the fast-growing artificial intelligence sector.
The open-weight model offers advanced AI agent capabilities, though it trails Kimi K3, developed by Chinese AI company Moonshot AI, in performance.
According to Times Now, V4-Flash costs about $0.14 per one million input tokens and $0.28 per one million output tokens. A token refers to a small unit of data processed by artificial intelligence models.
DeepSeek also cut token costs by 50 percent compared with previous pricing. The company halted plans for a previously announced dynamic pricing mechanism that would have doubled costs during peak usage hours.
The moves indicate DeepSeek is emphasizing its price advantage to expand its share of the artificial intelligence market.
According to testing by Artificial Analysis, V4-Flash costs an average of about 3 cents per task to run. Competing models cost significantly more: Reuters reported Kimi K3 costs about 86 cents per task, while OpenAI's GPT-5.6 Sol costs up to $1.86 per task.
The comparison accounts not only for base token prices but also for how many processing steps each model requires to complete a task. As a result, a model with lower token prices can become more costly overall if it needs more steps to finish a task.
DeepSeek's price cut comes as competition among Chinese technology companies grows increasingly intense.
The company drew significant global attention in early 2025 with its R1 model, which raised questions about the sustainability of heavy AI spending by American technology companies.
China's artificial intelligence race has expanded rapidly since then, with companies including Moonshot AI, MiniMax, Z.AI, ByteDance and Alibaba joining the competition.
According to Reuters, V4-Flash scored 50 out of 100 on the Intelligence Index compiled by Artificial Analysis, placing it at the same level as Google's Gemini 3.6 flash model. However, it trails Kimi K3 and some other leading global AI models.
DeepSeek is also developing a more advanced V4-Pro model alongside V4-Flash, though the company has not announced an official release timeline.
Set as preferred source
Set as preferred source
1 min read
1 min read
