The Chinese AI company DeepSeek is preparing a direct attack on OpenAI’s market leadership and is developing an AI agent system to be released by the end of 2025.
DeepSeek, founded by hedge fund manager Liang Wenfeng, is positioning itself for significant competition against the American AI giant OpenAI. The planned AI agent system will be able to perform multi-step actions on behalf of users, requiring minimal instructions. This development marks a strategic evolution from traditional chatbot interfaces to more autonomous and powerful AI systems.
DeepSeek’s timing strategy is particularly noteworthy as AI agents are considered the next big battleground in AI development. DeepSeek’s approach differs significantly from its competitors in both technical methodology and business philosophy. The company has already proven with its existing models that it can achieve competitive performance at a drastically reduced cost – DeepSeek-R1 was developed at a cost that is 90-95% lower than comparable models from OpenAI.
The multi-step action-capability approach that DeepSeek is developing presents a significant technical challenge. The system should be able to learn from its previous actions and continuously improve, which points to advanced memory systems and possibly reinforcement learning mechanisms.
Technical foundations and competitive position
DeepSeek’s technical foundation provides a solid platform for AI agent development. The latest DeepSeek-V3 model has 671 billion parameters in a Mixture of Experts (MoE) architecture with 37 billion activated parameters per token. This configuration demonstrates the company’s mastery of efficient large-scale model design.
The performance characteristics of the existing DeepSeek models provide insight into the expected capabilities of the AI agent system. DeepSeek-V3 operates at approximately 60 tokens per second – a threefold improvement over its predecessor DeepSeek-V2. The 128,000-token context window provides significant capacity for maintaining conversation histories and processing complex multi-step instructions.
Market dynamics and industry trends
The AI agent market represents one of the fastest growing segments within the AI industry. The global market size is expected to grow from $3.66 billion in 2023 to $139.12 billion by 2033 – a compound annual growth rate of 43.88%. This explosive growth trajectory provides the market context for DeepSeek’s strategic initiative.
Current market dynamics favor companies that can deliver powerful AI agent solutions at competitive prices. Enterprise adoption patterns show that 79% of executives report AI agent adoption in their organizations, but only 2% of companies have achieved full implementation. DeepSeek’s cost advantages and open source approach position the company to address some of the key barriers to broad AI agent adoption.
Key facts about the update
- Release schedule: DeepSeek plans AI agent system by the end of 2025
- Cost benefits: 90-95% lower development costs than OpenAI models
- Technical basis: DeepSeek-V3 with 671 billion parameters in MoE architecture
- Market size: AI agent market will grow from 3.66 billion to 139.12 billion dollars by 2033
- User base: 125 million monthly active users worldwide
- Funding: Independent funding from high-flyer hedge funds
- Open source model: Permissive licenses for developers and researchers
- Multi-step capabilities: Autonomous execution of complex multi-step tasks
- Web integration: First successful web search integration into reasoning models
- Pricing structure: Input cost of 0.27 dollars per million tokens
Source: Bloomberg
