deepseek · deepseek · chat & research
DeepSeek, with the working shown.
Low-cost reasoning for maths, logic and code, with the model's thinking visible before its answer.
- made by
- DeepSeek
- versions
- 2
- from
- 2 tokens
- included on
- Starter+
best at
What DeepSeek is good at
DeepSeek built its name on reasoning that costs very little. For maths problems, logic puzzles, data questions and code, it gets to careful answers at a price that makes it sensible to use for volume work, not just the occasional hard question.
DeepSeek V4 Pro thinks before it answers, and on LazySusan you can see that thinking: a collapsible reasoning block appears above the reply, so you can check how it reached a conclusion rather than taking the conclusion on trust. When the logic is wrong, you can usually see exactly where.
V4.1 Flash is the quick version, close enough to free that it suits drafts, rewrites and the questions you'd otherwise not bother asking. Both models can search the web when you switch Search on.
versions
DeepSeek versions on LazySusan
| Version | Use it for |
|---|---|
| DeepSeek V4.1 Flash | The default. Fast and efficient for everyday questions, drafting and code help. |
| DeepSeek V4 Pro | Thinks step by step before answering. For maths, logic, data analysis and tricky code. |
pricing
What DeepSeek costs on LazySusan
| Generation | Tokens |
|---|---|
| DeepSeek V4.1 Flash1–5 from a short to a long reply | 2 |
| DeepSeek V4 Pro4–17 from a short to a long reply | 7 |
Costs are for a message of about 300 characters; billing follows what each reply uses, including V4 Pro's thinking. A typical V4.1 Flash reply is 2 tokens, which makes DeepSeek one of the cheapest capable models on LazySusan.
Every plan includes DeepSeek. Starter is $8 a month with 1,000 tokens; unused tokens roll over while you're subscribed. Compare plans.
prompts
Prompts to try with DeepSeek
// maths
A tank fills in 6 hours with pipe A and 9 hours with pipe B, and drains in 12 hours. With all three open, how long does it take to fill? Show your working.
// data
Here's a CSV of monthly sales by region. Which region's growth is most consistent, and how did you measure consistency?
// code
Write a SQL query that returns each customer's second most recent order, and explain how it handles customers with only one order.
// logic check
Here is my argument for why we should delay the launch. Find any step where the conclusion doesn't follow from the premises.
// study
Explain Bayes' theorem using a medical-test example, then give me two practice problems with answers hidden at the end.
tips
Getting better results
- 01Read the reasoning block when the answer matters. It is the quickest way to spot a wrong assumption.
- 02Use V4.1 Flash for anything routine. At around 2 tokens a reply it costs a fraction of most models.
- 03Switch to V4 Pro for problems with a single correct answer, such as maths, logic or a stubborn bug. A typical V4 Pro reply is 7 tokens, thinking included.
- 04Ask DeepSeek to state its assumptions before solving. On word problems, most mistakes come from a misread condition rather than bad arithmetic.
- 05For a long chain of related questions, keep them in one conversation. DeepSeek reads the earlier messages each time, so it stays consistent with the definitions and numbers you've already agreed.
alternatives
When to pick a different tool
- Claude if you need polished writing or a careful read of a long document.
- ChatGPT if you want a broad generalist that can also make an image in the same thread.
- Perplexity if the answer depends on current information and you want sources.
- Multi-Chat if you'd like to check DeepSeek's answer against other models in one go.
also on the api
DeepSeek in your own app
Low-cost reasoning in your own app, with the same key and SDK as every other model.
One key, our SDKs for JavaScript and Python, prepaid credits with no subscription. LazySusan API · docs
faq
DeepSeek on LazySusan: questions
Yes. DeepSeek runs in the LazySusan workspace in any browser. It's included on every plan, from Starter at $8 a month.
DeepSeek V4.1 Flash for fast everyday use and DeepSeek V4 Pro for step-by-step reasoning.
Yes. When the model thinks before answering, its reasoning appears in a collapsible block above the reply.
A typical message costs about 2 tokens on V4.1 Flash and 7 tokens on V4 Pro. Long replies and attachments cost more.
It writes clearly and is fine for drafts, notes and explanations. For polished prose, a sensitive email or anything where tone matters, Claude or ChatGPT is usually the better pick, and switching is one click in the same workspace.
Your messages are sent to DeepSeek's API to generate the reply, as with any model you choose. LazySusan doesn't use your conversations to train models and doesn't sell your data.
Yes. DeepSeek V4.1 Flash and V4 Pro are in the developer API, on the same key and prepaid balance as every other model.
get started
DeepSeek and every other model, one plan.
One login, one bill and one Library for everything you make, with the right model for each part of the job.
Starter is $8/month. Cancel anytime.