DeepSeek to raise API prices up to 1,100% for V4 models

Chinese artificial intelligence startup DeepSeek announced Thursday that it will raise API prices for its V4-Pro and V4-Flash models, with increases ranging from 50% to 1,100% above current rates depending on the model, token type, and time of use. The new pricing, which introduces peak and off-peak tiers, will take effect on August 17.investing+1

From Preview Pricing to Commercial Reality

The price hike follows DeepSeek's launch of the V4-Pro-0813 model on August 12, which marked the flagship model's graduation from a nearly four-month preview period to general availability. During preview, DeepSeek had priced V4-Pro at $0.435 per million input tokens and $0.87 per million output tokens — a fraction of what competitors charge for comparable frontier models.wccftech+1

The company had telegraphed the increase. On August 6, DeepSeek indicated it planned a "relatively significant" price rise, though it did not disclose exact timing or scope at the time. Even after the increases take effect, analysts say the company retains substantial room for price adjustment while maintaining competitive capabilities against overseas rivals.globaltimes

Competitive Context

The pricing shift arrives amid intensifying competition in the AI model market. Elon Musk's xAI released Grok 4.6 on Wednesday, priced at $2 per million input tokens and $6 per million output tokens. OpenAI recently discounted its GPT-5.6 Luna model by as much as 80%, cutting output token prices to $1.20 per million.globaltimes+1

Even with substantial increases, DeepSeek's pricing would remain well below Western competitors. One comparison cited by market observers puts DeepSeek-V4-Pro output tokens at roughly $0.88 per million, compared with $50 per million for Anthropic's Fable 5.globaltimes

Surging Demand

The move reflects the commercial pressures facing DeepSeek as demand surges. The company was second only to Anthropic in total token consumption in July, according to industry data, and daily token usage across China surpassed 140 trillion in March — more than 1,000 times the level at the start of 2024.wccftech+1

V4-Pro-0813 is a mixture-of-experts system with 1.6 trillion total parameters and 49 billion active per token, offering a one-million-token context window. The model has drawn attention for approaching or matching leading Western models on agent-focused benchmarks at a fraction of the cost — what Tian Feng, former dean of SenseTime's Intelligence Industry Research Institute, described as a new phase where "the next competitive edge may depend less on who answers better in chat, and more on who can complete complex real-world tasks at lower cost and with higher reliability."globaltimes+1