| Category | Chat & assistants · Chatbots & LLMs | Chat & assistants · Chatbots & LLMs | Chat & assistants · Chatbots & LLMs |
|---|
| In one sentence | The Chinese AI that rattled Wall Street: free, open and dirt-cheap for developers, with privacy small print. | — | The search engine that answers with sources: ask as you would a person and check every fact in one click. |
|---|
| Free plan | ✓ Yes | ✓ Yes | ✓ Yes |
|---|
| Price | Free · API from $0.15/M tokens | Free | Free / Pro $20/mo / Max $200/mo |
|---|
| Plans and pricing | - Web chat and app €0
- API · deepseek-flash (V4.1-Flash) $0.30 / $1.20 per million tokens (input / output, peak)
- API · deepseek-v4-pro $1.32 / $3.96 per million tokens (peak)
- API · Search / Fetch $0.75 per 1,000 requests
| — | - Free €0
- Pro $20/month or $200/year (≈$16.67/month)
- Max $200/month or $2,000/year
- Education Pro $10/month
- Enterprise Pro / Max $40 / $325 per user per month
|
|---|
| AIAO score | 96/100 | 96/100 | 97/100 best |
|---|
| Languages | No verified data | No verified data | No verified data |
|---|
| Where to use it | Web, iOS, Android and API; DeepSeek Harness (developer preview). | No verified data | Web, iOS, Android, Windows and macOS; Comet browser; Sonar API. |
|---|
| Pros | - Free web chat and apps, no subscription.
- Open weights under the MIT licence: run it on your own server without sending anything to China.
- One of the cheapest APIs: Flash at $0.30 / $1.20 per million tokens at peak and half off-peak.
- 1M-token context and an API compatible with OpenAI and Anthropic formats.
| - Clasificada como TOP (96/100) por su adopción masiva en la industria, excelente calidad de resultados y fiabilidad superior frente a competidores directos.
| - Every answer links its sources so you can check them instantly.
- Up-to-date, real-time information.
- One subscription to choose between OpenAI, Anthropic and Google models.
- Comet, a free Chromium-based browser with an assistant that acts for you.
|
|---|
| Cons | - Sensitive privacy: data in China and no simple training toggle.
- Censorship on sensitive political topics.
- Regulatory blocks and complaints (Italy, Germany, consumer groups) and very fast model changes in the API.
| — | - Clashes with publishers (lawsuits from Dow Jones, The New York Times, Britannica, Reddit…).
- Limited free plan: about 3 Pro searches a day and 1 research report a month.
- Agentic-browser risks (prompt injection) and training on by default.
|
|---|
| Use cases | - Developers and companies wanting a powerful API at the lowest cost; students and curious users who want a free assistant for non-sensitive data; teams that prefer to deploy an open model on their own infrastructure.
- If you use the API for long jobs (summarising hundreds of documents, batch translation), schedule them off-peak and pay half. For personal or company data, run the open model locally or with a European provider.
| — | - Students, journalists and researchers who need sourced answers; professionals who want several frontier AIs with one subscription; anyone looking for a Google alternative for complex searches.
- Narrow the sources in the question itself (e.g. “only .gob.es sites and Spanish press from 2026”) and always open two or three citations before trusting a fact. If you use it for work, turn off “AI data retention” first.
|
|---|
| Privacy and data | Its policy (February 2026) says data is collected and stored in China and what you type may be used for training; opting out is by email and there is an EU representative. Alternative: its open models run locally or with a European provider. | No verified data | On Free, Pro and Max, “AI data retention” is on, so your queries may be used for training; turn it off in Settings > Preferences (only affects future data). On Enterprise, data is never used for training and files are deleted after 7 days. |
|---|
| Profile | View details → | View details → | View details → |
|---|
