DeepSeek-V4 Official Launch & Peak-Valley Pricing Explained

Introduction

In 2026, the LLM race delivers another major headline. On June 29, DeepSeek emailed users announcing that the DeepSeek V4 official release is planned for mid-July — moving this highly anticipated multimodal model from preview to full commercial availability.

For developers and enterprises worldwide, this is not only a technology upgrade but also a significant pricing shift. This article breaks down what's new in DeepSeek V4 official, how API pricing changes, and what it means for your business.

DeepSeek V4 Official: Beyond the Preview

As early as April 24, DeepSeek launched the V4 preview with Pro and Flash variants. The preview shipped with 1M-token context, thinking-mode toggles, JSON output, tool calling, prefix continuation, and other enterprise features across development, office, legal, and finance workflows.

The upcoming DeepSeek V4 official release builds on the preview with further optimizations. DeepSeek also introduced DSpark — a speculative decoding framework that boosts inference speed by 60%–85%, signaling better engineering efficiency and lower inference cost at launch.

API Pricing: Peak-Valley Mechanism Explained

DeepSeek API pricing is changing. After the official launch, the DeepSeek API platform will adopt peak-valley pricing — call prices double during peak hours.

Peak Hours

Daily 9:00–12:00 and 14:00–18:00 (Beijing time).

DeepSeek V4 Pro Pricing (per million tokens)

Billing Scenario	Off-Peak	Peak
Input (cache hit)	¥0.025	¥0.05
Input (cache miss)	¥3	¥6
Output	¥6	¥12

DeepSeek V4 Flash Pricing (per million tokens)

Billing Scenario	Off-Peak	Peak
Input (cache hit)	¥0.02	¥0.04
Input (cache miss)	¥1	¥2
Output	¥2	¥4

Why Peak-Valley Pricing?

Analysts view DeepSeek V4 peak-valley pricing not as a simple price hike but as a scheduling tool under scarce compute. V4 Flash alone exceeded 4.66 trillion tokens weekly for six consecutive weeks — topping global single-model usage.

Enterprise office-hour congestion and timeouts are now routine. Time-of-day pricing uses price signals to shift deferrable batch work off-peak, preserving stability for finance, coding, and real-time agents during business hours.

Pro vs Flash: How to Choose

When using V4 via the DeepSeek web app or DeepSeek API, pick the right variant for your scenario:

DeepSeek V4 Pro

• Stronger Agent capabilities; used internally as DeepSeek's Agentic Coding model
• Experience rivals Sonnet 4.5; delivery quality approaches Claude Opus 4.6 (non-thinking mode)
• Best for hard tasks and complex reasoning

DeepSeek V4 Flash

• Slightly less world knowledge than Pro, but comparable reasoning on many tasks
• Smaller parameters and activations — faster, more economical API
• Matches Pro on simpler workloads
• Ideal for cost-sensitive, latency-first scenarios

DeepSeek's Capital and Talent Push

DeepSeek funding news keeps flowing. On June 16, reports indicated a first external round exceeding ¥50 billion, with post-money valuation above ¥338 billion. Founder Liang Wenfeng contributed ~¥20B; Tencent ~¥10B; CATL ecosystem, NetEase, JD.com, and others participated.

DeepSeek is also hiring aggressively, planning to at least double headcount across algorithms, engineering, ops, and 33 roles in 7 categories.

The DeepSeek open platform is using capital and talent to close the commercialization gap with leading closed-source models overseas.

Summary

The imminent DeepSeek V4 official launch marks DeepSeek's shift from validation to full commercialization. Peak-valley pricing raises costs during busy hours but reflects high demand and a commitment to service stability.

Whether you use the DeepSeek web app or DeepSeek API platform for enterprise integration, understanding these changes helps you plan smarter. Schedule batch jobs off-peak for lower DeepSeek API costs.

DeepSeek is writing a new chapter in China's global LLM competition — with technical strength and commercial discipline.