DeepSeek raises API prices sharply as V4 Pro targets agentic AI market
DeepSeek said peak-hour billing will double some rates and raise output token costs by as much as 355%, according to its new schedule.
- On Sunday, August 16, 2026, Chinese artificial intelligence startup DeepSeek will officially transition its V4 Pro and V4 Flash models from flat-rate billing to dynamic peak and off-peak pricing at 16:00 UTC.
- The company designed this system to load-shape demand by incentivizing off-peak usage, which costs exactly 50% less than peak rates during the 17 off-peak hours of each day.
- Output token prices climb by up to 355% depending on tier and timing, forcing engineering teams that built pipelines around near-zero costs to re-evaluate their architectural budgets.
- At $8.78 peak and $4.39 off-peak on a 10M token workload, DeepSeek remains cheaper than Gemini 3.7 Flash and Claude 4.5 Haiku despite the price increases.
- DeepSeek CEO Liang Wenfeng operates an active compute fleet equivalent to approximately 20,000 top-tier GPUs, while existing users can optimize costs by batching workloads to the 17 off-peak hours.
33 Articles
33 Articles
DeepSeek V4 Pro Launches With Major Agent Upgrades and Open-Source Harness
DeepSeek has officially launched DeepSeek V4 Pro 0813, moving its 1.6-trillion-parameter flagship model out of preview and into general availability. The release closes a testing period that started in April and arrives alongside a new open-source agent framework and a significant API price increase. The model is now available in Expert Mode through the DeepSeek app and website, as well as through the API using the existing model names. DeepSeek…
The American Bazaar
DeepSeek launches V4 Pro at prices up to 14 times higher than V4 Flash
Chinese AI firm DeepSeek launched its V4 Pro model on Thursday. This model has been priced several times higher than its previous V4 Flash model. DeepSeek’s V4-Pro-0813 is priced at $1.32 per million input tokens and $3.96 per million output tokens, according to independent benchmarking firm Artificial Analysis. This is the equivalent of around nine times the input price and 14 times the output price of V4 Flash, which Artificial Analysis liste…
DeepSeek, a Chinese artificial intelligence (AI) darling, released on Thursday the official version of its model of AI V4 Pro, presenting very improved agent capabilities, but at a price for the user substantially higher than previously announced by the company. Exclusive material for subscribers. To have full access, access the link of the material and register.
Tech Brief (Aug. 14): DeepSeek Launches V4-Pro and Raises API Prices by as Much as 1,100%
Tech Brief (Aug. 14): DeepSeek Launches V4-Pro and Raises API Prices by as Much as 1,100% - WeRide targets break-even by 2029 with driver-assistance sales, asset-light overseas push, Honor’s new smartphone doubles as a robotic camera operator
DeepSeek Price Increase: Rates and DeepSeek Alternatives
Developers and enterprise teams worldwide were taken by surprise when Chinese artificial intelligence lab DeepSeek published a formal update to its commercial platform. Effective August 16, 2026, at 16:00 UTC, the company is retiring its legendary flat-rate billing structure for its flagship DeepSeek V4 Pro and DeepSeek V4 Flash models in favor of a dynamic peak and off-peak pricing system. With output token costs climbing by up to 355% and peak…
Coverage Details
Bias Distribution
- 38% of the sources lean Left, 37% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
