Loading...
Loading...
Facts checked against both vendors on October 13, 2026.
Choose ChatGPT if you want a finished, hosted product with tiers from 0 to 100 US dollars a month and above, plus business plans with admin controls and regional data options. Choose DeepSeek if token price or open weights decide the purchase: its API lists 0.30 US dollars per million cache-miss input tokens and 1.20 US dollars per million output tokens for Flash at peak, and its V4-Pro model card states the weights are MIT licensed, so you can run them yourself. If you use DeepSeek's own hosted app or API, its privacy policy says personal data is processed and stored in the People's Republic of China, and that fact should be settled before price is compared.
ChatGPT and DeepSeek are not the same kind of purchase. ChatGPT is a closed product from OpenAI: you rent access to models through a hosted app and a metered API. DeepSeek publishes model weights that can be downloaded and run elsewhere, and also sells access to the same family of models through its own app, web chat and API at much lower listed token prices. The decision therefore turns on three things: price per token, whether you need the weights, and where your data goes. Figures below come from each vendor's own pages as read on 13 October 2026. OpenAI's main pricing and help pages refused automated access, so ChatGPT plan details come from an OpenAI-run documentation page that is focused on its Codex coding tool, and are flagged where that matters.
ChatGPT is sold as a service. OpenAI's pricing documentation lists Free at 0 US dollars a month, Go at 8 US dollars, Plus at 20 US dollars, and Pro from 100 US dollars, with Pro tiers at 100, 200 and 500 US dollars a month. Business is listed at 20 dollars per user per month when billed annually for two or more users, or 25 dollars per user per month billed monthly. Enterprise and Edu are contact-sales. Only the Pro tiers state USD explicitly on that page; the other prices show a dollar sign without a currency code.
DeepSeek's V4-Pro model card on Hugging Face says: "This repository and the model weights are licensed under the MIT License." MIT is a permissive licence, so the weights can in principle be hosted on your own hardware or by a third-party provider. The same card lists V4-Pro at 1.6 trillion parameters with 49 billion activated, and V4-Flash at 284 billion parameters with 13 billion activated. The card was the only weights page read for this comparison; the licence on any other DeepSeek release should be checked on its own page, since an earlier release, V3, pointed to a separate Model License for its weights.
OpenAI's models are not published as weights, so there is no self-hosting option for ChatGPT's models. If running a model inside your own environment is a requirement, that alone decides the question.
DeepSeek's API documentation lists two models, deepseek-flash and deepseek-v4-pro, each with a 1 million token context length and a maximum output of 384,000 tokens. Prices are in US dollars per million tokens. Flash costs 0.15 for cache-miss input, 0.003 for cache-hit input and 0.60 for output at off-peak rates, and double those at peak: 0.30, 0.006 and 1.20. V4-Pro costs 0.66, 0.022 and 1.98 off-peak, and 1.32, 0.044 and 3.96 at peak.
The off-peak discount is a flat half. DeepSeek defines peak as 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays. Every other hour, including weekends and Chinese public holidays, is off-peak.
OpenAI's API pricing page lists several current models. At standard rates, per million tokens in US dollars: gpt-6-sol is 2.00 input, 0.20 cached input and 10.00 output; gpt-6.1-sol is 2.00, 0.10 and 10.00; gpt-6-luna is 0.10, 0.01 and 0.50; gpt-6-astra is 10.00, 1.00 and 50.00. Prompts above 272,000 input tokens are billed at higher long-context rates. OpenAI's gpt-6-sol model page lists a 1,050,000 token context window and 128,000 maximum output tokens. Batch and Flex rates are about half of standard.
The price gap is not uniform. Take a monthly job of 10 million cache-miss input tokens and 2 million output tokens. On DeepSeek Flash that is about 5.40 US dollars at peak rates or 2.70 off-peak; on V4-Pro about 21 dollars at peak or 10.50 off-peak. On gpt-6-sol it is 40 dollars. But on OpenAI's smaller gpt-6-luna it is 2.00 dollars, which is below DeepSeek Flash even at off-peak rates on these listed figures. DeepSeek is far cheaper against OpenAI's mid-tier models, and not cheaper against its smallest one. Output quality at each price point has to be tested on your own material.
DeepSeek's privacy policy states: "To provide you with our services, we directly collect, process and store your Personal Data in People's Republic of China." It says account-linked data is kept for as long as you have an account, with longer retention possible in some cases, and that personal data is used to "train and improve our technology, such as our machine learning models and algorithms." The policy describes a right to opt out of using personal data for training. DeepSeek's API terms of service are governed by the laws of the People's Republic of China in the mainland, with disputes going to a court with jurisdiction over the registered office of Hangzhou DeepSeek Artificial Intelligence Co., Ltd. Those terms do not state where servers are located and do not say whether API inputs are used for DeepSeek's own training.
OpenAI's API documentation says data sent to the API is not used to train its models unless you opt in. Abuse monitoring logs are kept for up to 30 days by default, and a Zero Data Retention option exists for eligible endpoints. Data residency is a project setting: regional processing is listed for the United States, Europe (the EEA plus Switzerland) and the United Arab Emirates, and regional storage only for Australia, Canada, Japan, India, Singapore, South Korea and the United Kingdom. Regional processing endpoints carry a 10 percent price uplift for models released on or after 5 March 2026. OpenAI's pricing documentation also lists retention and residency controls under Enterprise and Edu. OpenAI's consumer privacy policy could not be fetched, so consumer ChatGPT data handling is not described here.
The open weights change the picture for DeepSeek models. The processing location in the privacy policy applies to DeepSeek's own services. If the weights are run on infrastructure you choose, the data path is whatever you build.
ChatGPT's tiers scale by usage. On the pricing documentation read here, Plus and standard Business limits are counted per five-hour period, with weekly limits possible, and Pro has no five-hour limit. Extra usage can be bought as credits on Plus and Pro. That page is written around Codex, OpenAI's coding tool, so it does not give a full consumer feature list for the chat app. Its own table is also not fully consistent: the Plus feature row names GPT-6.1 Sol and GPT-6 Luna, while the limits table includes rows for GPT-6 Astra and GPT-6 Sol on Plus.
DeepSeek's home page offers an app download by QR code, a web chat and links to its API platform and documentation. It does not state a price for the app or web chat, and the API documentation page does not either, so whether the consumer apps are free could not be confirmed from vendor pages.
DeepSeek's documentation says its API format is compatible with OpenAI and Anthropic formats, and that OpenAI or Anthropic SDKs can be pointed at it by changing configuration. That lowers the cost of testing DeepSeek against an existing OpenAI integration: the same prompts can be run on both and compared on quality, latency and price. It does not guarantee that every feature of one provider's API exists on the other. DeepSeek also notes that the legacy model names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted and are served by DeepSeek-V4.1-Flash at Flash pricing.
Built from each tool's listing. Vendors can update their own listing after claiming it.
ChatGPT lists Free at 0 US dollars a month, Go at 8, Plus at 20, and Pro from 100 (tiers at 100, 200 and 500 US dollars a month). Business is 20 dollars per user per month billed annually for two or more users, or 25 monthly. Enterprise and Edu are contact-sales. API prices are per million tokens, for example gpt-6-sol at 2.00 input and 10.00 output in US dollars.
DeepSeek's API lists Flash at 0.30 US dollars input (cache miss) and 1.20 output per million tokens at peak, and half that off-peak. V4-Pro is 1.32 and 3.96 at peak, 0.66 and 1.98 off-peak. The app and web chat prices are not stated on the pages read.
One email on Tuesday with what we researched that week. No spam, unsubscribe anytime.
