DeepSeek has announced a significant price increase for its APIs, possibly due to a surge in demand following its low-price strategy.



On July 31, 2026, DeepSeek, a Chinese AI company, released its AI model '

DeepSeek-V4-Flash-0731 ' as an open model under a commercially usable license. DeepSeek-V4-Flash-0731 was announced as an even cheaper model than OpenAI's low-cost model ' GPT-5.6 Luna ,' but the significant price reduction to compete led to such high demand that the infrastructure couldn't keep up, forcing DeepSeek to raise prices.

DeepSeek Plans 'Significant' Price Increase for AI Services - Bloomberg
https://www.bloomberg.com/news/articles/2026-08-06/deepseek-plans-significant-price-increase-for-its-ai-services

DeepSeek Forced To Raise Prices As Its Recent Price Cuts To Snub OpenAI Unleashed A Demand Tsnunami That Its 20,000-GPU Stash Can't Handle
https://wccftech.com/deepseek-forced-to-raise-prices-as-its-recent-price-cuts-to-snub-openai-unleashed-a-demand-tsnunami-that-its-20000-gpu-stash-cant-handle/

DeepSeek's $8 Billion Fundraising Reopens, Valuation Near $74 Billion - Bloomberg
https://www.bloomberg.com/news/articles/2026-08-06/deepseek-resumes-8-billion-round-with-monolith-in-the-running

OpenAI's AI model 'GPT-5.6,' officially released on July 9, 2026, is divided into three types: the flagship model 'GPT-5.6 Sol,' the balanced model for everyday tasks 'GPT-5.6 Terra,' and the low-cost 'GPT-5.6 Luna.' It is emphasized that even the low-cost GPT-5.6 Luna has performance comparable to the previous generation model, GPT-5.5. The API price for GPT-5.6 Luna is $1 (approximately 162 yen) per 1 million input tokens and $6 (approximately 971 yen) per 1 million output tokens, which is one-fifth the price of GPT-5.6 Sol and two-fifths the price of GPT-5.6 Terra.

OpenAI officially releases 'GPT-5.6,' demonstrating coding performance exceeding Claude Fable in third-party tests - GIGAZINE



Then, on July 31, 2026, OpenAI announced a significant price reduction for the GPT-5.6 series . GPT-5.6 Terra saw a 20% price reduction, while GPT-5.6 Luna was reduced by a massive 80%, now costing $0.20 (approximately 31 yen) per million input tokens and $1.20 (approximately 190 yen) per million output tokens.




This significant price reduction has made GPT-5.6 Luna more cost-effective than not only closed models like Claude Sonnet 5 and Gemini 3.6 Flash, but also Chinese open models like DeepSeek V4 Pro and GLM-5.2. OpenAI explained that they were able to achieve this significant price reduction by 'extracting further architectural efficiencies from the model,' but some point out that this is 'OpenAI's pricing strategy aimed at Chinese AI research institutes.'

Then, just a few hours after OpenAI's price reduction, Chinese AI company DeepSeek released its AI model 'DeepSeek-V4-Flash-0731'. DeepSeek-V4-Flash-0731 was touted as an extremely cost-effective model, costing $0.14 (approximately 22 yen) per 1 million input tokens and $0.28 (approximately 44 yen) per 1 million output tokens, which is even cheaper than the reduced price of GPT-5.6 Luna.

'DeepSeek-V4-Flash-0731' appears, an open model offering comparable performance to GPT-5.6 Luna at a lower price - GIGAZINE



However, on August 6, 2026, DeepSeek notified users of a significant price increase for DeepSeek-V4-Flash-0731. The specific amount of the price increase was not stated in the notification, but DeepSeek urged users to plan their usage in light of the price increase.



The reason for the price increase has not been explained, but some say that 'DeepSeek is responding to an unprecedented surge in demand.' There have also been reports of the model's inference speed occasionally becoming extremely slow due to the surge in demand. The following is the official DeepSeek API status log, which shows that access difficulties occurred frequently before and after the release of DeepSeek-V4-Flash-0731.



DeepSeek CEO Liang Wenfeng revealed that as of July 2026, 'DeepSeek currently possesses computing resources equivalent to 20,000 NVIDIA H100 GPUs,' suggesting that the surge in demand due to their low-price strategy may have strained those computing resources.

Furthermore, according to Bloomberg, DeepSeek is in the midst of a large-scale fundraising campaign and may hold an initial public offering (IPO) as early as 2026. Trading was temporarily suspended at the end of July 2026 following the leak of a statement attributed to CEO Wen Feng explaining the US-China AI competition to investors, but it was reported on August 6, 2026, that the company had resumed its second round of fundraising.

DeepSeek is aiming to raise approximately $8 billion (approximately 1.2 trillion yen) in funding, and its valuation after funding is expected to reach approximately $74 billion (approximately 11 trillion yen). DeepSeek is said to be struggling to balance investor expectations, rapid market expansion, and the enormous capital requirements necessary to build its expensive computing infrastructure.

Related Posts:

in AI, Posted by log1e_dh