DeepSeek review: our take
DeepSeek has some of the lowest published API prices for a model with a 1M-token context: deepseek-flash costs $0.30 per million input tokens and $1.20 per million output tokens at peak, and half that off-peak. Peak means 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, so weekends and US daytime run at the lower rate. The chat app is free with no consumer paid plan, and the weights are MIT-licensed, so you can also run them yourself.
The trade-off is data. DeepSeek's privacy policy says personal data is collected, processed and stored in China and used for training unless you opt out. If that rules it out, Mistral Small 4 is listed at a similar $0.15 input and $0.60 output, and OpenRouter lets you test several low-cost models behind one key. Our OpenAI API alternatives page compares providers by price.
By the Prompt Hackers team · facts checked 4 Oct 2026
What DeepSeek does
DeepSeek's chat runs on the web and in iOS and Android apps, with Instant, Expert and Vision modes and photo and file uploads. Its API serves deepseek-flash (DeepSeek-V4.1-Flash) and deepseek-v4-pro with a 1M-token context and accepts both the OpenAI and the Anthropic request formats. Off-peak API rates are half the peak rates, and the model weights are published under the MIT license.
Who DeepSeek is for
Best for
- developers who want very low per-token prices
- users who want a capable chatbot with no subscription
- teams that can self-host open weights
Not for
- anyone who cannot store data in China: the privacy policy says personal data is stored there
- users who expect training to be opt-in: inputs are used for training unless you opt out
DeepSeek features
- F1Free chat app
- F2Instant, Expert and Vision modes
- F31M-token context API
- F4Peak and off-peak pricing
- F5OpenAI- and Anthropic-format API
- F6MIT-licensed open weights
- F7DeepSeek Harness desktop agent
DeepSeek pricing and plans
| Plan | Monthly | Billed yearly, per month | What it includes |
|---|---|---|---|
| Free chat app | $0 | — | Web, iOS and Android; no ads or in-app purchases at launch; no consumer paid plan; no published message limit |
| deepseek-flash, peak (per 1M tokens) | — | — | input $0.30 (cache hit $0.006) / output $1.20; 1M context, 384K max output, vision |
| deepseek-flash, off-peak (per 1M tokens) | — | — | input $0.15 (cache hit $0.003) / output $0.60 |
| deepseek-v4-pro, peak (per 1M tokens) | — | — | input $1.32 (cache hit $0.044) / output $3.96; 1M context, no vision |
| deepseek-v4-pro, off-peak (per 1M tokens) | — | — | input $0.66 (cache hit $0.022) / output $1.98 |
Free plan: yes. Prices as listed by the vendor in USD, before tax; they change, so check the pricing page before you buy.
DeepSeek FAQ
- How much does DeepSeek cost?
- DeepSeek charges by usage; see the pricing table above.
- Does DeepSeek have a free plan?
- Yes, DeepSeek has a free plan.
- What are the alternatives to DeepSeek?
- Tools that cover the same jobs include Mistral Vibe, OpenRouter, ChatGPT, Claude.
Sources
- api-docs.deepseek.com/quick_start/pricing
- api-docs.deepseek.com/updates
- deepseek.com/en/news/deepseek-v4-1-flash
- api-docs.deepseek.com/news/news250115
- deepseek.com/en/harness
- deepseek.com/en
- apps.apple.com/us/app/deepseek-ai-assistant/id6737597349
- cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html
- huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash
- huggingface.co/deepseek-ai/DeepSeek-V4-Pro
- openrouter.ai/api/v1/models
- trustpilot.com/review/deepseek.com