DeepSeek V4-Flash Becomes the Lowest-Cost Well-Known AI Model as Chinese AI Competition Heats Up
Table of Contents
According to research firm Artificial Analysis, DeepSeek V4-Flash is currently the lowest-cost well-known AI model evaluated by the firm, with an estimated average cost of $0.03 per benchmark task. The findings highlight how Chinese AI companies are rapidly narrowing the performance gap with Western rivals while competing aggressively on pricing.
The report comes as several Chinese AI developers continue releasing increasingly capable models, intensifying competition across the global AI industry.
DeepSeek V4-Flash Sets a New Pricing Benchmark
Artificial Analysis estimated that DeepSeek V4-Flash costs approximately 3 cents per benchmark task, making it more than 60 times cheaper than OpenAI’s GPT-5.6 Sol, which averages $1.86 per task, and over 100 times cheaper than Anthropic’s Claude Fable 5, which averages $3.15 per task.
Unlike headline token pricing, the benchmark cost-per-task measures the amount of data a model processes and generates to complete typical workloads, providing a more practical comparison of operating costs.
DeepSeek prices DeepSeek V4-Flash at $0.14 per million input tokens and $0.28 per million output tokens. The model entered public beta on 31 July.
Competitive Performance at Lower Cost
Despite its low operating cost, DeepSeek V4-Flash delivers competitive benchmark performance.
Artificial Analysis awarded the model 50 out of 100 on its Intelligence Index, matching Google’s Gemini 3.6 Flash and finishing just one point behind Meta’s Muse Spark 1.1.
The model is built on a 284-billion-parameter mixture-of-experts (MoE) architecture, with only 13 billion parameters active per token, helping reduce computational requirements while maintaining strong performance.
Alibaba Launches Qwen3.8-Max
The same day the Artificial Analysis report was published, Alibaba introduced Qwen3.8-Max, its largest AI model to date.
According to Alibaba, the model features 2.4 trillion parameters, supports a one-million-token context window, and ranked as the highest-performing Chinese model on the Arena.AI text leaderboard while placing second globally on the vision leaderboard behind Anthropic’s Fable 5 variants.
Reuters also reported that Alibaba’s Hong Kong-listed shares rose 7% following the announcement.
China’s AI Competition Continues to Intensify
Competition among Chinese AI developers continues to accelerate, with companies including ByteDance, Moonshot AI, MiniMax, Z.AI, Alibaba, and DeepSeek introducing new AI models and expanding their capabilities.
Reuters also reported that DeepSeek is preparing for a potential initial public offering (IPO) while continuing development of V4-Pro, a more advanced successor to DeepSeek V4-Flash. However, the company has not announced a launch date for the upcoming model.
Lian Jye Su, chief analyst at research firm Omdia, told Reuters that many Chinese AI companies are focusing on building models that are “good enough, affordable, transparent and accessible” rather than simply pursuing the highest benchmark scores.
Key Highlights
| Features | Details |
| Model | DeepSeek V4-Flash |
| Average Benchmark Cost | Approximately $0.03 per task |
| Token Pricing | $0.14 input / $0.28 output per million tokens |
| Intelligence Index | 50/100 |
| Architecture | 284B MoE with 13B active parameters |
| Major Competitor Launch | Alibaba Qwen3.8-Max |
| Future Plans | DeepSeek V4-Pro in development; potential IPO reported by Reuters |
Conclusion
DeepSeek V4-Flash demonstrates how Chinese AI developers are combining competitive performance with significantly lower operating costs. According to Artificial Analysis, the model currently offers the lowest benchmark cost among well-known AI models it has evaluated, while Alibaba and other Chinese firms continue to introduce increasingly capable alternatives. As competition intensifies, affordability is becoming as important as raw model performance in shaping the next phase of the global AI market.
