DeepSeek peak/off-peak pricing update
DeepSeek rolls out its DeepSeek-V4-Pro model, boasting major agent upgrades and flexible reasoning capabilities for various task complexities. Crucially, they're also revamping their API pricing with a novel peak/off-peak structure that slashes off-peak rates by 50%. This combination of advanced AI and dynamic economic incentives is designed to grab the attention of developers looking for both power and cost efficiency.
The Lowdown
DeepSeek has officially released its DeepSeek-V4-Pro model, bringing advanced AI capabilities to general availability, coupled with an intriguing new API pricing strategy. This update aims to enhance AI agent performance while offering developers more cost-effective options through a future-dated, time-based billing system.
- The DeepSeek-V4-Pro model is now generally available, accessible via their app/web in "Expert Mode" and through their API.
- It features substantial "Agent" upgrades, promising improved performance for production workflows.
- Users can leverage a "flexible reasoning effort" (thinking_mode) for V4-Pro and V4-Flash, allowing customization for simple, daily agent, or complex tasks.
- The API supports native OpenAI Responses API, specifically optimized for easy integration with Codex.
- A significant pricing update introduces distinct peak and off-peak rates for API usage, with off-peak costing 50% less.
- These new pricing tiers are slated to take effect on August 16, 2026.
DeepSeek's announcement pairs a more powerful AI model with a forward-looking, time-sensitive pricing scheme, signaling a move towards more granular control over both computational resources and developer expenditure.