From Time-Limited Discount to Permanent DeepSeek API Pricing Reset
DeepSeek has converted what looked like a short-term promotion into a permanent reset of its DeepSeek API pricing. The company’s V4-Pro model will now stay at one quarter of its original list price, rather than snapping back after the discounted period. On the current pricing page, DeepSeek-V4-Pro is listed at USD 0.435 (approx. RM2.00) per million uncached input tokens and USD 0.87 (approx. RM4.00) per million output tokens, down from crossed-out reference prices of USD 1.74 (approx. RM8.00) and USD 3.48 (approx. RM16.00). Cached input falls to USD 0.003625 (approx. RM0.02) per million tokens after the same 75 percent reduction. The lighter V4-Flash model is even cheaper at USD 0.14 (approx. RM0.64) per million input tokens and USD 0.28 (approx. RM1.30) per million output tokens, with cache hits at USD 0.0028 (approx. RM0.01). In practical terms, the sale price has simply become the real price.
How a 75 Percent Cut Changes Developer API Savings and Product Design
By locking in a 75 percent reduction, DeepSeek has turned AI API costs into a strategic lever for developers and founders. AI-native products often suffer from infrastructure-like cost structures: every support reply, coding task, or research query generates a token bill that can silently erode margins. With DeepSeek-V4-Pro now ranging from USD 0.003625 (approx. RM0.02) to USD 0.87 (approx. RM4.00) per million tokens, and additional cuts bringing cache-hit pricing across the lineup down to one tenth of launch levels, the economics of experimentation look different. Low-priced or freemium features become more realistic, especially for tools aimed at students, small businesses, or solo professionals. Teams can afford to keep more context in prompts instead of aggressively compressing documents just to save tokens. In short, developer API savings move from a constraint to an enabler, reshaping what LLM-powered products can viably offer.
LLM Pricing Comparison: Undercutting Premium Rivals to Win Market Share
DeepSeek’s move is as much about competitive positioning as it is about generosity. The company is explicitly leaning into the “cost-effective” label for large-context agents, with V4 models positioned as ushering in a new era of affordable 1M context length. Against premium offerings like GPT-5 or Gemini 3.5 Flash, DeepSeek’s price ladder—V4-Pro for heavier reasoning, V4-Flash for everyday workloads—creates a stark LLM pricing comparison. For enterprises and power users processing millions of tokens daily, the gap between USD 0.14 (approx. RM0.64) and USD 0.435 (approx. RM2.00) per million input tokens versus higher-priced incumbents presents a compelling alternative. Even if not every organization switches, procurement teams now have a benchmark to challenge existing contracts, and engineering teams can architect routing systems that reserve expensive models only for edge cases, while DeepSeek handles the bulk of routine traffic.
Risks, Trade-Offs and the Next Phase of AI API Cost Competition
The permanent discount also signals that DeepSeek is willing to trade margin for reach. Frontier models remain expensive to train and run, yet DeepSeek is promising aggressive pricing while reportedly emphasizing breakthrough research and ongoing open-source releases. That strategy assumes scale and architectural efficiency will eventually offset thinner per-token margins. For developers, low AI API costs are attractive but not the whole story. Reliability, latency, data handling, tool calling capabilities, and trust all influence provider choice. Some enterprises may hesitate to place their most sensitive workloads on a newer platform, even with steep savings. Still, the price delta is large enough that many teams will at least integrate DeepSeek as a secondary or fallback model. This is likely to accelerate a broader shift where model providers compete less on flashy demos and more on the day-to-day economics of shipping profitable AI products.