Chinese AI startup DeepSeek has launched DeepSeek-V3.1, an updated version of its large language model.

The company said the new model features a hybrid inference structure and improvements in processing speed and agent capabilities.

DeepSeek also plans to update its API pricing for developers starting September 6.

The company is based in China and develops AI models for integration into apps and web products.

🔗 Source: Reuters

🧠 Food for thought

1️⃣ Cost-efficient AI development emerges as key competitive advantage

DeepSeek’s V3.1 demonstrates how Chinese AI companies are achieving comparable performance at dramatically lower costs than their US counterparts.

The model was trained for just $5.6 million and achieves a 71.6% score on the Aider coding benchmark, rivaling proprietary systems from OpenAI and Anthropic12.

This cost efficiency stems partly from US chip export restrictions, which pushed Chinese firms to optimize their models for maximum performance per dollar rather than simply scaling with more hardware3.

The approach contrasts with the resource-intensive development strategies of American AI companies, suggesting that innovation under constraints can drive more efficient solutions.

DeepSeek’s open-source strategy further enhances this advantage by eliminating the need to recoup massive development costs through premium pricing, creating a sustainable competitive edge based on operational efficiency rather than just technical capability.

2️⃣ API pricing wars intensify as models become commoditized

DeepSeek’s upcoming API cost adjustments reflect a broader trend of aggressive pricing competition that’s reshaping the AI market.

This follows OpenAI’s recent 80% price reduction for its O3 API, bringing costs down to $2 per million input tokens and $8 per million output tokens4.

The shift toward usage-based API pricing models allows developers to pay only for what they use, making advanced AI capabilities more accessible to smaller companies and startups5.

Recent DeepSeek developments