DeepSeek V4 Pro Raises Prices to Reroute AI Demand
💡DeepSeek’s 350% price hike reveals where inference capacity is breaking—and how AI teams should schedule workloads.
⚡ 30-Second TL;DR
What Changed
Peak-hour V4 Pro output pricing rose from 6 to 27 yuan per million tokens, while cache-hit input pricing rose from 0.025 to 0.30 yuan.
Why It Matters
AI application teams may face materially higher real-time inference costs and will need to schedule batch workloads during cheaper periods. For infrastructure vendors, higher model monetization improves the outlook for longer-term capacity commitments and usage-linked pricing.
What To Do Next
Benchmark your DeepSeek V4 Pro workloads by peak and off-peak windows, then move batch inference and cache-warming jobs to discounted periods.
Key Points
- •Peak-hour V4 Pro output pricing rose from 6 to 27 yuan per million tokens, while cache-hit input pricing rose from 0.025 to 0.30 yuan.
- •DeepSeek V4 Flash usage reached 8.83 trillion weekly tokens, up 570% week over week, while the API began returning capacity-shortage errors.
- •Off-peak periods are priced at half rate, and weekends are charged entirely at the low-usage rate to shift batch workloads away from daytime peaks.
- •The article links model pricing to stronger AI infrastructure demand, including optical modules, foundries, equipment, and cloud capital expenditure.
- •Compute providers are beginning to explore token-revenue-linked contracts instead of fixed hourly or monthly GPU rentals.
🧠 Deep Insight
Background and context from public sources — not the original article. 11 sources cited.
🔑 Enhanced Key Takeaways
- •DeepSeek's pricing restructure, effective August 16, 2026, was framed as a discount mechanism despite the off-peak rates still exceeding the previous flat-rate pricing model.
- •The company reported 475 million RMB in revenue for the first seven months of 2026, against a massive 11 billion RMB investment in compute infrastructure, highlighting a significant burn rate.
- •DeepSeek is currently in late-stage negotiations for a 50 billion RMB funding round, targeting a pre-money valuation of approximately 500 billion RMB.
- •The price hikes are partially attributed to hardware constraints resulting from US export controls, which force reliance on older, less efficient AI accelerators to meet surging demand.
- •DeepSeek-V4-Pro (version 0813) reached General Availability on August 13, 2026, introducing native support for the OpenAI Responses API and enhanced agentic capabilities.
📊 Competitor Analysis▸ Show
| Feature | DeepSeek V4 Pro | Qwen (Alibaba) | GLM (Zhipu AI) |
|---|---|---|---|
| Pricing Model | Dynamic Peak/Off-Peak | Tiered/Volume-based | Tiered/Volume-based |
| Primary Focus | Agentic/Coding | General Purpose/Cloud | Enterprise/B2B |
| API Compatibility | OpenAI Responses API | OpenAI Compatible | OpenAI Compatible |
🛠️ Technical Deep Dive
- DeepSeek-V4-Pro (0813) architecture includes native support for the OpenAI Responses API to facilitate easier developer migration.
- The model incorporates advanced agentic capabilities designed for complex reasoning and multi-step task execution.
- The experimental deepseek-v4-flash-vision-exp model, released August 21, 2026, utilizes a specialized multimodal architecture for enhanced visual-agent performance.
- Infrastructure reliance involves older-generation hardware clusters due to restricted access to cutting-edge US-designed accelerators.
🔮 Future ImplicationsAI analysis grounded in cited sources
⏳ Timeline
📎 Sources (11)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 虎嗅 ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
Weekly AI briefing
One email a week. Unsubscribe anytime.