Chinese AI startup DeepSeek has launched a preview of its V4 large language model, focusing on agent-based tasks and inference. The open-source release continues the company's trend of challenging high-cost US AI development models.
- V4 preview focuses on agent-based tasks and knowledge processing
- Maintains open-source distribution for local modification
- Offers 'pro' and 'flash' size variants
- Compatible with Claude Code and OpenClaw
- Challenges the high-cost infrastructure model of US AI labs
Sign up free to read the full analysis
Create a free account to unlock full AI-curated market articles, personalized alerts, and more.