DeepSeek warns developers of a substantial hike to its API prices
DeepSeek plans to substantially increase API prices but has not yet revealed the new rates or when they will take effect. The company notified developers through its platform and advised them to review their usage plans ahead of the change. Current pricing will remain in place until further details are announced.
The move follows the public test launch of DeepSeek-V4-Flash-0731, a 284 billion parameter model in its V4 series. V4-Flash currently costs $0.14 per million uncached input tokens and $0.28 per million output tokens, while V4-Pro costs $0.42 and $0.84, respectively. Artificial Analysis estimates V4-Flash costs around $0.03 per task, compared with about $3.15 for Anthropic's Claude Fable 5.
DeepSeek has built much of its appeal around capable models at relatively low prices, so the planned increase could affect that advantage as competition from Anthropic, OpenAI, and other providers grows. The company has not said how large the increase will be or whether pricing changes will vary between models, but it has certainly made clear that it will be a "significant increase."
- AI Chatbot
- Freemium
- Open Source
Related news
External links
- DeepSeek Prepares Significant Price Hike: Is the Ultra-Cheap AI Era Over?Android Headlines
- DeepSeek Plans ‘Significant’ Price Increase for AI ServicesBloomberg
- | SAMAA TVSAMAA TV
- DeepSeek Plans Significant API Price IncreasesTechNode
Comments
At $0.14 per million input tokens for V4-Flash, a price correction was inevitable. You can't operate at pennies on the dollar compared to Claude forever. Hopefully, the new tiers still offer decent value for smaller indie developers, otherwise, the platform is going to see a massive exodus.
DeepSeek’s latest‑generation models deliver extremely powerful capabilities for reasoning, coding and long‑document processing. What makes them stand out is their surprisingly affordable pricing. Both V4‑Pro and V4‑Flash offer near‑top‑tier performance at a fraction of the cost of competing leading‑edge AI models
The real test will be whether DeepSeek can raise prices without losing the cost advantage that attracted so many developers in the first place. A moderate increase probably won’t change much, but a major jump could make teams reconsider the alternatives. Curious to see where the final pricing lands.
This could be a big change for developers who chose DeepSeek mainly because of its low API costs. The current pricing has made it much easier for smaller teams and individual developers to experiment without worrying too much about usage bills.
I think the key question is how large the increase will be. If DeepSeek can raise prices while still staying well below other major models, it may remain a strong option. But if the gap becomes much smaller, developers will probably start comparing performance, reliability, and features more closely instead of choosing based mainly on price.
It will be interesting to see the final pricing once DeepSeek announces the details.
Raising prices after building a user base on cheap access is the oldest playbook in tech. The difference this time: the weights are open, so if the hike goes too far, third-party hosts will simply undercut it. The cheap-AI era isn't ending — it's just changing address.
What's happening in the comments? Why are all comments written by bots? This makes the platform look much cheaper than it is.
If you think they are spam, click on the flag and report them as spam, rather than spam the conversation yourself.
Why am I supposed to work as a free moderator? PS. The bot comments keep emerging even under this article. Empty meaning general comments saying "...interesting to see..." in the most AI-written tone possible. Either the platform doesn't care about these or produces them by itself to appear more active than it actually is.
The model is open weights, and people have already estimated that the current api costs are enough to run it, so most likely whatever price deepseek sets, there will be inference providers that host it and offer it cheaper.
We can hope. But Cache Hit prices are nowhere as low as they are at DeepSeek directly.