Skip to content

Chat & searchDeepSeek

DeepSeek

DeepSeek is a free AI chat assistant from the Chinese lab DeepSeek, built on its open-weight V4 model family. It pairs a no-subscription assistant with very low per-token API prices, and DeepSeek-V4.1-Flash arrived in September 2026. Budget-minded users, developers and researchers who self-host models use it most.

Quick facts

VendorDeepSeek
CategoryChat & search
PlatformsWebiOSAndroidAPI
Free planYes
Pricing fromNot listed
Last reviewed

Best for

  • Budget-conscious everyday users
  • Developers cutting API costs
  • Researchers using open weights

Key features

Free web and mobile chat

Chat, upload files and search the web at chat.deepseek.com or in the iOS and Android apps without a subscription.

1M-token context

Since the V4 release in April 2026, a 1M-token context window is the default across DeepSeek's official services.

V4-Pro in Expert Mode

DeepSeek-V4-Pro, generally available since August 13, 2026, is offered in the app and on the web through Expert Mode.

Open weights

Weights and technical reports for V4-Pro, V4-Flash and V4.1-Flash are published on Hugging Face for research and self-hosting.

Off-peak API pricing

Per-token billing with context caching, and off-peak rates that cost half the price of weekday peak windows.

Works with coding agents

The API accepts both OpenAI and Anthropic request formats, so tools like Claude Code and OpenCode can run on DeepSeek models.

Use cases

  • Debug a script or walk through a math proof in the free web chat without paying for a plan.
  • Point Claude Code or OpenCode at DeepSeek's Anthropic-compatible endpoint to run coding agents on cheaper tokens.
  • Schedule large batch summarization jobs outside peak UTC hours to halve the API bill.
  • Download open weights from Hugging Face to test a DeepSeek model on your own infrastructure.

Pricing

The web chat and iOS and Android apps are free. The API is pay-as-you-go: deepseek-flash costs $0.30 per 1M input tokens (cache miss) and $1.20 per 1M output tokens at peak, half that off-peak.

Ways to save

  • Run flexible API workloads outside peak hours (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday) to pay 50% less.
  • Reuse long prompt prefixes: cached deepseek-flash input costs $0.006 per 1M tokens at peak instead of $0.30 on a cache miss.
  • Pick deepseek-flash over deepseek-v4-pro for routine jobs; its peak output price is $1.20 per 1M tokens versus $3.96.
Prices checked on Oct 12, 2026

Current deals

No live deals right now — check back or see all deals.AI Deals

What is DeepSeek?

DeepSeek is both a Hangzhou-based AI lab and the free assistant it runs in the browser and in mobile apps. Developers reach the same models through a low-cost API or download the open weights to run themselves.

Recent changes

  • April 24, 2026: V4 Preview introduced V4-Pro (1.6T total, 49B active parameters) and V4-Flash, and made 1M context standard.
  • July 24, 2026: the legacy deepseek-chat and deepseek-reasoner API model names were retired.
  • August 2026: V4-Pro reached general availability, and peak and off-peak API rates took effect.
  • September 10, 2026: V4.1-Flash replaced V4-Flash with cheaper API pricing.
  • UK: the service now carries the name DSeek.

Limitations

DeepSeek stores personal data in China, which may rule it out for regulated or confidential work. In the API, vision input works only with deepseek-flash, and prices double during weekday peak windows. Model names and routing have changed several times in 2026, so API users should check the pricing page before deploying.

DeepSeek FAQ

Yes. The DeepSeek web chat and its iOS and Android apps cost nothing and have no subscription tier. Only the developer API is paid, deducted per token from a topped-up or granted balance.

In China. DeepSeek's privacy policy, last updated February 10, 2026, says it collects, processes and stores personal data in the People's Republic of China. Weigh that before sharing sensitive personal or company information.

A notice on chat.deepseek.com says that, due to brand restructuring, DeepSeek is now officially DSeek in the UK. The notice adds that all services continue as normal.

It's DeepSeek's newest model, released September 10, 2026: a 552B-parameter mixture-of-experts model with native visual understanding. It replaced V4-Flash in the API under the name deepseek-flash, with lower prices.

Yes. DeepSeek first planned to route V4-Pro requests to V4.1-Flash from September 14, 2026, but its pricing page now says V4-Pro stays available with unchanged billing until further notice.

Articles about DeepSeek

Workshop pegboard wall filled with neatly hung hand tools and power tools
Guides

Best Free AI Tools 2026: 20 Free Plans Worth Using, by Category

As of October 2026, the best free AI tools are ChatGPT, Gemini and DeepSeek for chat, GitHub Copilot for coding, Canva and Freepik for images, Kling and CapCut for video, ElevenLabs and Suno for audio, and NotebookLM for research. Every free plan has a cap, usually credits that reset daily or monthly.

10 min read

Close-up of hands typing a message on a smartphone
Comparisons

ChatGPT vs Claude vs Gemini (2026): Prices, Free Plans and Which to Pick

ChatGPT, Claude and Gemini all have free plans, and their main paid tiers cost $20, $20 and $19.99 a month. Pick ChatGPT for an all-rounder with Codex on every plan, Claude for long documents and Claude Code, and Gemini for Google apps, video and 5 TB of storage. US students can get ChatGPT Plus or Google AI Pro free for months.

8 min read

Sources

Some links are affiliate links. We may earn a commission at no extra cost to you — it never changes which deals we list. Affiliate disclosure