Chinese artificial intelligence startup DeepSeek has launched V4-Flash, an ultra-low-cost AI model that is rapidly drawing attention across the industry for combining competitive performance with some of the lowest operating costs currently available in the market.
The launch is being viewed as the latest move in an increasingly intense global AI price war, as companies race to deliver advanced AI capabilities while driving down costs for developers and businesses. According to research firm Artificial Analysis, DeepSeek’s V4-Flash is now the cheapest well-known AI model to run among major commercial offerings.
Lowest-Cost Model Among Major AI Providers
Artificial Analysis reported that DeepSeek charges $0.14 per million input tokens and $0.28 per million output tokens for V4-Flash. The firm’s benchmark testing estimated an average evaluation cost of roughly $0.03 per run, significantly below the costs associated with several competing frontier AI models.
The pricing structure is confirmed in DeepSeek’s official API documentation, which lists V4-Flash as the company’s most economical flagship model.
The release reinforces DeepSeek’s reputation for focusing on cost-efficient AI systems, a strategy that previously helped the company gain international attention and challenge larger rivals in the rapidly evolving AI sector.
Related
- DeepSeek Unveils Latest Models One Year After Disrupting Global Tech
- DeepSeek Outage Halts Millions, Raises Concerns Over AI Reliance
Performance Remains Competitive
While V4-Flash is positioned as a budget-friendly model, industry evaluations indicate that it remains highly competitive in terms of capabilities.
Artificial Analysis assigned the model an intelligence score of 50 on its benchmark index, placing it alongside Google’s Gemini Flash-class systems and close to several higher-priced competitors. The combination of strong performance and extremely low cost is what has made the launch notable across the AI industry.
DeepSeek describes V4-Flash as a model designed for speed, efficiency and affordability. In its official announcement, the company said the model’s reasoning capabilities “closely approach V4-Pro” while offering faster responses and lower operating costs.
The company also stated that V4-Flash performs on par with its larger V4-Pro model for simpler agent-based tasks.
Built for Large-Scale AI Workloads
According to DeepSeek, V4-Flash features a context window of up to 1 million tokens, allowing it to process large documents and extended conversations. The model supports both standard chat interactions and more advanced agent-based workflows.
DeepSeek’s official release announcement described V4-Flash as:
“Your fast, efficient, and economical choice.”
The company said the model supports both thinking and non-thinking modes and is compatible with OpenAI-style and Anthropic-style API integrations.
Competition Intensifies
The launch comes at a time of heightened competition among Chinese AI developers. Companies including Alibaba, Moonshot AI, MiniMax, ByteDance and Zhipu AI have all introduced increasingly capable models as they compete for market share in both domestic and international markets.
On the same day, Alibaba unveiled Qwen3.8-Max, its largest and most capable AI model to date, underscoring the rapid pace of innovation taking place within China’s AI industry.
Analysts say DeepSeek’s latest release could place additional pressure on AI providers worldwide to lower prices while maintaining high performance standards. As AI adoption expands across software development, enterprise automation, customer service and content generation, operating costs are becoming an increasingly important factor for businesses selecting AI models.
DeepSeek originally introduced the V4 model family in April and recently expanded availability of the official V4-Flash API through a public beta release. The company has indicated that a full release of its more powerful V4-Pro model is expected to follow.
