Skip to main content
DeepSeek comes in two main lines — a flagship chat model and a reasoner (the reasoning, R-series model). Choosing between them comes down to one question: do you want a fast all-rounder for everyday work, or step-by-step reasoning for a hard problem. The reason people reach for DeepSeek is simple — strong capability at a very low price. This page helps you pick a model for the task; the technical side of connecting is in the Quickstart.

​
How the lines differ

LineStrong atSpeed / costReach for it when
Flagship chatAll-rounder: code, text, general tasks, long contextFaster, very cheapEveryday work, code, bulk processing
ReasonerStep-by-step reasoning: math, logic, hard problemsSlower, pricier than flagshipWhen you need the task worked through step by step, not just answered
Not sure? Start with the flagship chat. It covers most work and costs next to nothing. Bring in the reasoner selectively, when the task genuinely needs step-by-step reasoning: hard math, logic, untangling a tricky problem.

​
What DeepSeek can do

  • Long context — the window fits long documents and large chunks of code.
  • Strong code and reasoning — the flagship writes and edits code confidently; the reasoner adds step-by-step working for hard problems.
  • Very low price — DeepSeek’s key advantage: quality close to top models at a noticeably lower per-token cost.
  • Function calling (tool use) — the flagship chat can return structured tool calls; don’t rely on this for the reasoner. See function calling for a full example.
For long answers, you can also stream the response token by token instead of waiting for the full reply.
DeepSeek models are text-only: you can’t pass images as input, there’s no vision. If you need to read images, look at Claude or Gemini instead.

​
Picking by task

  • Cheap and high volume → flagship chat, for price and speed.
  • Hard math, logic, step-by-step working → reasoner.
  • Writing and refactoring code → flagship chat.
  • Agent or tool pipeline → flagship chat (it has reliable tool use).
  • Long documents → flagship chat with long context.

​
How to connect

DeepSeek is served over the OpenAI-compatible protocol — no separate library needed, take any OpenAI SDK and change the base_url to https://www.ruapi.ai/v1. Step-by-step setup and a first request are in the Quickstart.
The model field takes an ID, e.g. deepseek-v3 for the flagship chat and deepseek-reasoner for reasoning. Exact names, versions and per-token prices are on the Pricing page at www.ruapi.ai.

​
FAQ

The flagship chat (V3) is the everyday pick: code, text, general tasks, and very cheap on top. Use the reasoner when you specifically need step-by-step reasoning — hard math, logic, untangling a tricky problem.
DeepSeek is one of the best-value models per token: capability close to the top tier at a noticeably lower price. Exact prices are on the main site’s Pricing page.
No. DeepSeek models are text-only and don’t take images as input. For images, use Claude or Gemini with vision.
Large — the window fits long documents and big chunks of code. The exact limit depends on the version and matches DeepSeek’s official one.
You create a key in your dashboard after signing up. How to wire it into your code is in the Quickstart.
Top up in USDT — with crypto. How to add balance is on the Top-up page.

​
Next

Connect and first request

base_url, key, Python and curl code.

Claude models

Same key — Claude’s lineup and capabilities.