{"slug": "deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels", "title": "DeepSeek raises peak prices of flagship AI models to more than four times current levels ahead of IPO", "summary": "DeepSeek will raise peak-hour prices for its flagship V4 models by more than four times from current levels starting Aug 16, with the DeepSeek-V4-Flash model priced at US$1.32 per million output tokens during peak hours, up from US$0.28, and the V4-Pro at US$3.96, up from US$0.87. The Chinese AI lab, which is preparing for a potential IPO as soon as 2026, said the dynamic pricing aims to allocate resources more reasonably and encourage off-peak usage, while its rates remain far below rivals like Moonshot's Kimi K3 at US$15 and Anthropic's Fable 5 at US$50 per million output tokens.", "body_md": "# DeepSeek raises peak prices of flagship AI models to more than four times current levels ahead of IPO\n\nIncrease suggests firm is placing more focus on profitability as it eyes a stock-market debut\n\nDEEPSEEK is steeply raising the prices for its flagship V4 models ahead of a potential initial public offering, even as the low-cost provider’s rates remain well below those of its major artificial intelligence rivals.\n\nA new peak-hour pricing will increase the Chinese company’s rates by more than four times from the current levels, according to a post on its website on Thursday (Aug 13). The price increases will take effect on Aug 16.\n\nGoing forward, users will pay US$1.32 for 1 million output tokens during peak hours, and half that during off-peak hours, for the DeepSeek-V4-Flash model. That is up from US$0.28 for 1 million tokens previously. A token is a [measure of the work an AI model does](https://www.businesstimes.com.sg/companies-markets/telcos-media-tech/end-tokenmaxxing-how-ai-moving-pay-you-go) to digest information and produce results.\n\nThe increase suggests DeepSeek is placing more focus on profitability as it eyes a stock-market debut, though the move still leaves its cost per token below that of main competitors. Chinese rival Moonshot prices its Kimi K3 at US$15 per million output tokens, while Anthropic’s state-of-the-art Fable 5 service charges US$50.\n\nDeepSeek’s V4-Pro model will cost US$3.96 for 1 million tokens at peak hours and half that at non-peak hours, it said. That is up from the current US$0.87 per million tokens.\n\nThe Hangzhou-based AI lab said it is revising and adjusting the pricing “to allocate resources more reasonably.” The dynamic pricing strategy is designed to encourage developers and enterprises to shift their work to less congested periods. DeepSeek gave an advance warning on Aug 6 that it would hike prices, without specifying the exact increases.\n\nDeepSeek’s strategy has sparked a global debate in recent weeks over the [cost of high-performing AI tools.](https://www.businesstimes.com.sg/startups-tech/technology/openai-meta-spacexai-compete-more-cost-efficient-ai-models) Its pricing has been so far out of the usual range for AI models that it has prompted the idea of a DeepSeek “death zone,” where costlier or less capable models would be obviated. It has also raised questions about the IPO plans of OpenAI and Anthropic because of the risks the Chinese firm poses to their profits and business models.\n\nDeepSeek is in the middle of a massive fundraising and has [begun preparations for an IPO as soon as in 2026,](https://www.businesstimes.com.sg/startups-tech/technology/chinas-deepseek-raise-fresh-capital-us74-billion-valuation-ahead-onshore-ipo) Bloomberg has reported. Founder Liang Wenfeng will for the first time need to balance investor expectations, rapid market expansion and the intense capital demands of building out costly computing infrastructure.\n\nThe Chinese startup also announced on Thursday a developer preview of its DeepSeek Harness v0.1, signalling it is preparing to take on services such as Anthropic’s Claude Code. The term harness refers to the software scaffolding that wraps around a large language model that enables AI agents to manage more complex tasks – such as read files, edit code, browse the web and work its way through a task until done.\n\nSEE ALSO\n\nInstead of hard-coding the product, as some US rivals do, DeepSeek said its open architecture allows users to plug in any component of choice, including other AI models, to the harness. That strategy aims to improve its own agent ecosystem. BLOOMBERG\n\nDecoding Asia newsletter: your guide to navigating Asia in a new global order. Sign up here to get Decoding Asia newsletter. Delivered to your inbox. Free.\n\nShare with us your feedback on BT's products and services\n\n[TRENDING NOW](/pulse?ref=trending-now)\n\nThe S$1 million National Day paradox: ‘money dysmorphia’ amid a wealth surge\n\nSingapore at 61: How we can ensure opportunity, security and ownership for the next generation\n\nWhy Asean matters more than ever to the UK and Singapore\n\nToo little, too late? Manila’s billion-dollar bid to ignite its sputtering EV industry", "url": "https://wpnews.pro/news/deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels", "canonical_source": "https://www.businesstimes.com.sg/companies-markets/telcos-media-tech/deepseek-raises-peak-prices-flagship-ai-models-more-four-times-current-levels-ahead-ipo", "published_at": "2026-08-14 00:45:16+00:00", "updated_at": "2026-08-14 01:07:09.780975+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-products", "ai-policy"], "entities": ["DeepSeek", "DeepSeek-V4-Flash", "DeepSeek-V4-Pro", "Moonshot", "Kimi K3", "Anthropic", "Fable 5", "Liang Wenfeng"], "alternates": {"html": "https://wpnews.pro/news/deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels", "markdown": "https://wpnews.pro/news/deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels.md", "text": "https://wpnews.pro/news/deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels.txt", "jsonld": "https://wpnews.pro/news/deepseek-raises-peak-prices-of-flagship-ai-models-to-more-than-four-times-levels.jsonld"}}