Technology

DeepSeek's New AI Models Are Here — and So Are Higher, Time-Based Prices

Martin HollowayPublished 4month ago4 min readBased on 6 sources
Reading level
DeepSeek's New AI Models Are Here — and So Are Higher, Time-Based Prices
source:deepseek.com

DeepSeek has launched its newest AI models, V4-Pro and V4-Flash, along with a new pricing system that charges different rates depending on the time of day. The new prices take effect at 16:00 UTC on August 16, 2026 (DeepSeek API Docs).

The V4 lineup includes two models. V4-Pro is available through the DeepSeek app and website via a feature called "Expert Mode," and also through DeepSeek's API, which is the service developers use to build applications on top of DeepSeek's models. Both V4-Pro and V4-Flash let developers choose how hard the model thinks about a request: a light setting for simple tasks, a higher setting for everyday work, and a maximum setting for complex problems. V4-Pro also works directly with OpenAI's API format, making it easier for developers already using OpenAI's tools to switch over.

The new pricing splits costs into peak and off-peak windows. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, per DeepSeek's pricing documentation (DeepSeek API Docs). Off-peak rates are 50% lower than peak rates. For V4-Flash, peak pricing is $1.32 per 1 million output tokens, with off-peak at half that amount (PYMNTS). The overall changes represent an increase ranging from 50% to 1,100% above current prices (Investing.com).

This is not DeepSeek's first time trying time-based pricing. The company previously offered off-peak discounts on its V3.1 model that ended on September 5, 2025 at 16:00 UTC, after which standard pricing took effect (DeepSeek API Docs). An earlier preview version of V4 was open-sourced in April 2026 with the ability to process up to 1 million tokens of text at once, setting the stage for this full release (DeepSeek API Docs).

The broader context here is that time-based pricing is common in other infrastructure businesses. Electricity grids, cloud computing, and internet bandwidth services have all charged more during busy hours to smooth out demand. DeepSeek is applying the same idea to AI model usage, because the graphics processors that run these models are not equally busy at all hours of the day. The company is passing that cost difference directly to its customers rather than averaging it into one flat price. Whether other AI providers like OpenAI, Anthropic, and Google follow this approach will depend on how they each manage their own computing capacity as demand grows.

The range of the increase — 50% at the low end to 1,100% at the high — suggests DeepSeek is raising prices on certain types of usage quite aggressively. For developers who can run their workloads during off-peak hours, the 50% discount makes a meaningful difference. For those whose applications need to respond instantly and cannot wait for cheaper hours, the cost increase could be significant.

The three reasoning effort levels give developers a separate way to manage costs: they can choose how much computing power to spend on each request. Combined with the time-based pricing, this means developers have two levers to control what they spend — how hard the model works on each task, and when those tasks run. V4-Pro's compatibility with OpenAI's API and its optimization for Codex, OpenAI's coding tool, further reduce the effort required for developers already using OpenAI's platform to make the switch.

For teams currently using DeepSeek's API, the time to prepare is short. The new rates activate at 16:00 UTC on August 16, giving developers roughly 48 hours from the August 14 announcement to review their usage patterns, identify which models and token types they consume, and estimate the cost impact under both peak and off-peak scenarios. Model names remain unchanged, so no code changes are required for existing API calls — but the billing impact will be immediate.