DeepSeek-V3.2-Exp Launches with Major Price Cut
Silicon-Based Flow Unveils DeepSeek-V3.2-Exp with Significant Improvements
Silicon-based Flow has announced the launch of DeepSeek-V3.2-Exp, its latest experimental AI model featuring groundbreaking improvements in context handling and cost efficiency. The new version supports an impressive 160K context length while reducing prices by over 50% compared to previous iterations.
Technical Advancements
The model builds upon DeepSeek-V3.1-Terminus architecture with optimized long-text processing capabilities. Key to this improvement is the implementation of DeepSeek Sparse Attention, a fine-grained attention mechanism that maintains benchmark performance while dramatically enhancing long-context understanding.
Platform Integration
As the 140th model on Silicon-based Flow's platform, DeepSeek-V3.2-Exp joins a growing ecosystem accessible through multiple interfaces including:
- Model Square
- Claude Code
- Gen-CLI
- Cline
- Trae
The company emphasizes that while the experimental version shows promise, they will continue supporting V3.1-Terminus API services for production environments requiring absolute stability.
Developer Benefits
The platform offers several advantages for AI developers:
- Unified access to all models with single authentication
- Free trial credits (¥14/$1 for new users)
- Model comparison tools for performance evaluation
Silicon-based Flow maintains its commitment to providing diverse AI solutions across language, image, audio, and video domains through its expanding model library.
Key Points:
🌟 Enhanced Capacity: Supports unprecedented 160K context length
💰 Cost Reduction: Pricing decreased by over 50%
🔍 Diverse Options: Platform now hosts 140 specialized models
