Chinese startup DeepSeek is causing fresh disruption in the AI market with the release of V4-Flash. According to data from research firm Artificial Analysis, the new model has emerged as the world's most economical, with an average cost of just $0.03 per test.
The Staggering Cost Gap
Comparisons with leading American models are revealing. The operating cost of V4-Flash is more than 100 times lower than Anthropic's Claude Fable 5, which costs $3.15 per test. Similarly, OpenAI's GPT-5.6 Sol stands at $1.86.
DeepSeek charges $0.14 per million input tokens and $0.28 per million output tokens. Artificial Analysis emphasizes that measuring cost per test is more representative than simple token pricing, as it accounts for the steps and iterations required to complete a task.
Performance vs. Price
Despite its ultra-low cost, V4-Flash lags in raw power. It scored 50 out of 100 on the Intelligence Index, matching Google's Gemini 3.6 Flash but falling significantly behind Moonshot AI's Kimi K3 (57 points) and top-tier models from OpenAI and Anthropic.
- V4-Flash is just one point behind Meta's Muse Spark 1.1.
- DeepSeek is already preparing a more powerful version, V4-Pro.
- Alibaba also introduced Qwen3.8-Max, intensifying domestic Chinese competition.
DeepSeek, which triggered international market volatility in early 2025 with its R1 model, is now reportedly considering an IPO, seeking to capitalize on its high-efficiency, low-cost strategy.
