What happened
Chinese AI developer DeepSeek moved its flagship model, V4 Pro, out of preview and into general availability on August 13, designating the build DeepSeek-V4-Pro-0813 after a preview period that ran nearly four months. The release adds adjustable reasoning effort, letting developers dial the model between low for simple tasks, high for daily agent workflows, and max for complex problems, plus native support for the OpenAI Responses API with one-click setup for Codex-style coding tools. The bigger news is pricing. Starting August 16 at 16:00 UTC, DeepSeek replaces its flat per-token rates with peak and off-peak pricing, with off-peak rates set 50 percent below peak, and the company has signaled that meaningful price increases are coming. Reuters coverage noted that V4 Pro is priced at up to 14 times the cost of the company's cheaper V4 Flash model, a striking move for a firm that built its reputation on undercutting Western AI labs.
Why it matters for your business
Even if you have never heard of DeepSeek, there is a fair chance some inexpensive AI tool in your stack runs on it, because its rock-bottom prices made it the default engine for budget AI products. When the cheapest supplier in a market starts raising prices and segmenting by time of day, the products built on it eventually pass costs along, or quietly swap to a cheaper, possibly weaker model behind the scenes. Either outcome affects what you are paying for. The off-peak discount is also a genuinely useful pattern to know about: heavy batch work like document processing or report generation can be scheduled into cheaper windows, the same way businesses already think about electricity.
What to do about it
- Ask your AI tool vendors which models power their product and how pricing changes get passed through.
- If your business uses AI APIs directly, schedule bulk jobs into off-peak windows where the provider offers them.
- Avoid hard-wiring any single model into a workflow; the market is moving too fast for lock-in to be safe.