Updated

DeepSeek

Low-cost API for DeepSeek's open-weight reasoning and coding models, compatible with the OpenAI and Anthropic API formats.

11 makers' products we track use DeepSeek. For LLM API it's 7th most chosen of 12 options: 11 of the 923 makers' products we know the LLM API of (1%). Among the 256 launched in 2026, 0% chose it, against 1% of those launched in 2024 or earlier. 25 open-source projects also declare it in their code. What makers like most: It delivers strong reasoning at a small fraction of frontier model prices. Their most common complaint: It burns many thinking tokens, so cost per finished task can exceed cheap per-token prices. On GitHub, 68 public pull requests moved projects to DeepSeek from tools we list, and 48 moved them away.

No free tierPaid from Pay per token

Profile checked

Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

How makers use DeepSeek

Makers' products we track use DeepSeek for LLM API (11), most often alongside Next.js (45%).

What DeepSeek is used for

LLM API · Model provider →Low per-token prices on strong reasoning and coding models, with OpenAI- and Anthropic-format endpoints.Trade-off: Served from China, which matters for some customers' data requirements; peak-hour prices are double off-peak.

DeepSeek vs the alternatives makers chose

For LLM API, 11 of the 923 makers' products we know the choice of picked DeepSeek (1%), behind OpenAI API (80%).

OptionMakers' products2026 launchesFree forAt 10,000 usersOpen source
OpenAI API736 80%77%~1,000 users$10/mo · GPT-6 LunaNoCompare →
Claude252 27%21%~1,000 users$10/mo · Claude Haiku 5.5NoCompare →
Gemini API99 11%12%~360 users$28/mo · Gemini 3.1 Flash-LiteNoCompare →
Groq39 4%3%~1,500 users$6.75/mo · GPT OSS 20BNo
Azure OpenAI22 2%0%——No
Mistral AI15 2%1%~1,700 users$6.00/mo · Ministral 3 (3B)NoCompare →
DeepSeek11 1%0%~370 users$27/mo · deepseek-flashNo

For LLM API: how many of the 923 makers' products we know the LLM API of chose each option, and the share among the 256 launched in 2026, since a tool that came first has products from every year, and what each costs at 10,000 users.

Costs are the cheapest plan that fits, from each vendor's pricing page, at the usage our stack planner assumes for a product with that many users. Free for: the users its cheapest plan a business can use stays free for.

Often paired with

Share of the 11 makers' products on DeepSeek that show this choice.

When to use DeepSeek, and when not

Pick DeepSeek if

  • Token cost dominates your budget, for example high-volume batch or agent workloads. LLM API →

Something else fits better if

  • You want the strongest general models and the widest ecosystem of examples and integrations: OpenAI API or Claude or Gemini API.
  • You want to try or mix many models with one key and one bill, and fall back when a provider is down: OpenRouter.
  • You want that routing and spend control but self-hosted, with keys and logs on your servers: LiteLLM.
From each decision's guide, written from what makers' products chose and why.

DeepSeek is an AI lab based in Hangzhou, China. Its API, which DeepSeek calls the Open Platform, serves its current models: a fast, low-cost Flash model and a larger Pro model, both with a 1M-token context window, tool calls and JSON output.

More about DeepSeek

There is no DeepSeek-specific SDK to learn. The API speaks the OpenAI Chat Completions format at https://api.deepseek.com, the Anthropic Messages format at /anthropic, and the Responses API, so you use the official OpenAI or Anthropic SDKs, or agent tools such as Claude Code, by swapping the base URL and key. Thinking mode is on by default with an adjustable effort level and can be turned off per request. Usage is deducted from a prepaid balance.

Useful for a small team: context caching on disk is on for everyone, so repeated prompt prefixes are billed at the cache-hit rate without code changes; limits are counted as concurrent requests per account rather than requests per minute, and DeepSeek raises them on request at no cost; a user_id field isolates your own end users for caching and scheduling; and the models are published as open weights on Hugging Face, so the same models can be hosted elsewhere.

The main limit is jurisdiction: DeepSeek's privacy policy says the personal data it collects is processed and stored in the People's Republic of China, and the platform terms are governed by PRC law. Anthropic compatibility is partial too, with fields such as cache_control and mcp_servers ignored and document blocks unsupported.

DeepSeek pros and cons

What makers like most: It delivers strong reasoning at a small fraction of frontier model prices. Their most common complaint: It burns many thinking tokens, so cost per finished task can exceed cheap per-token prices.

Pros

  • It delivers strong reasoning at a small fraction of frontier model prices. PHPH 2HNHN 2
  • Recent Flash models handle real coding work in agent harnesses close to frontier models. HNHN 2HN 3HN 4
  • It has fewer guardrails, so it helps with security and malware analysis other models refuse. HNHN 2HN 3HN 4
  • The direct API gets very high cache hit rates, keeping heavy use cheap. HNHN 2
Product Hunt review tags
deep thinking ability11
supports coding tasks10
handles complex tasks7
fast performance6
open source6
free tier5

Cons and watch-outs

  • It burns many thinking tokens, so cost per finished task can exceed cheap per-token prices. HNHN 2HN 3HN 4
  • It hallucinates, ignores instructions and struggles with end-to-end builds that frontier models finish. HNHN 2HN 3HN 4
Product Hunt review tags
slow performance5
not as good as ChatGPT2
server busy errors2
times out2

Themes that recur in makers' words and Hacker News comments, each linked to what it summarises, with how Product Hunt tags its reviews.

4.9★ from 55 reviews on Product Hunt ↗

What Hacker News says about DeepSeek

On Hacker News, 33% of comments about DeepSeek recommend it and 10% criticize it: net praise +23%, 6th best liked of 12 tools HN discusses for LLM API. Commenters chose it 3.6× as often per month in 2026 as in 2025.

By year

YearComments choosing it, per monthNet praise
20230.2—
20241.3+29%
202515+21%
2026 (8 mo)56+24%

In their words

  • “DeepSeek flash has been killing it since I switched it on” HN, Aug 2026
  • “I am usually really happy with DeepSeek v4 Flash that I feel just mostly does what I tell it to do” HN, Jul 2026
  • “the Deepseek v4 flash model I am using through OpenRouter is killing me on latency” HN, Jul 2026

Hacker News comments January 2023 to August 2026 that name DeepSeek in a sentence like "I use" or "we moved to", Claude read a sample of 77 and the counts scale it to all of them; net praise is comments recommending it minus comments criticizing it, as a share of all comments about it. How it's counted, and how HN compares with makers.

Who switches

More projects move to DeepSeek: 68 pull requests in, 48 out; the most come from Gemini API and leave for OpenAI API.

Moving to DeepSeekMoving away
DeepSeek
Gemini API → DeepSeek24 PRs
OpenAI API → DeepSeek21 PRs
Claude → DeepSeek16 PRs
Groq → DeepSeek7 PRs
DeepSeek → OpenAI API16 PRs
DeepSeek → Gemini API15 PRs
DeepSeek → Groq9 PRs
DeepSeek → Claude8 PRs

Public pull requests on GitHub since Oct 2024 whose title says "X to Y" — real code changes moving a project from one tool to another, by developers in general.

Who uses it

11 makers' products, each linked to the source that shows it, and 25 open-source projects that declare it in their code. The best known on Product Hunt: MGX (Now Atoms) (3,165 followers), Genstore.ai (1,812 followers), Blink.new (1,524 followers), IFTTT (703 followers), Webhound (692 followers).

The maker says so 9Subprocessor list 2Declared in code 25How evidence is collected →
Browser Use
Browser UseOpen-source library and cloud service that lets AI agents control a browser to complete…

“Best open source model currently. Gives Browser Use superpowers!”

LLM APIMaker says so · source ↗
IFTTT
IFTTTEvery thing works better together

“DeepSeek delivers serious reasoning at a fraction of the cost of comparable models. We chose it because powerful AI shouldn't have a premium price tag attached to every use case.”

LLM APIMaker says so · source ↗
Weavable
WeavableGive every AI agent persistent work context

“DeepSeek gives us frontier-level reasoning at a fraction of the cost, which means we can run AI features at scale without watching the meter. Perfect for the high-volume inference paths where every token counts.”

LLM APIMaker says so · source ↗
MGX (Now Atoms)
MGX (Now Atoms)The World's First AI Dev Team

“DeepSeek Chat serves as our reliable foundation for basic content creation tasks. Its efficient performance and consistent output quality make it perfect for handling straightforward development requests.”

LLM APIMaker says so · source ↗
Captioner
CaptionerAI subtitle generator for video content creators.

“I use this Chinese LLM to provide natural sounding Chinese subtitles for the videos.”

LLM APIMaker says so · source ↗

Open source: a project that declares DeepSeek as a dependency in its public code — verifiable, but not necessarily a live product.

What its projects tell coding agents

16 of the 22 open-source products built with DeepSeek that were worked on in the last 180 days keep instruction files for coding agents — 73%, against 53% of all the projects we read. AGENTS.md is in 13, CLAUDE.md in 8.

1,043words in the median file
pnpmthe package manager most named, in 3 of 15 files
67%say how to test a change
Named more in these files than in other projects' — in how many of the 15:
What AGENTS.md files say about other tools →What open-source products tell their coding agents →

Alternatives to DeepSeek

All alternatives by situation →

Further reading

Questions makers ask about DeepSeek

Is DeepSeek good for production?

11 makers' products we track run on it. On GitHub, 68 public pull requests moved projects to DeepSeek from tools we list, and 48 moved them away. The watch-out makers raise most: It burns many thinking tokens, so cost per finished task can exceed cheap per-token prices.

What is DeepSeek used for?

Low-cost API for DeepSeek's open-weight reasoning and coding models, compatible with the OpenAI and Anthropic API formats. Makers' products we track use it for LLM API (11).

What are the pros and cons of DeepSeek?

Pros makers name: It delivers strong reasoning at a small fraction of frontier model prices. Recent Flash models handle real coding work in agent harnesses close to frontier models. It has fewer guardrails, so it helps with security and malware analysis other models refuse. Cons: It burns many thinking tokens, so cost per finished task can exceed cheap per-token prices. It hallucinates, ignores instructions and struggles with end-to-end builds that frontier models finish.

What are the alternatives to DeepSeek?

For LLM API, the other options makers' products we track chose most are OpenAI API (736), Claude (252), Gemini API (99), Groq (39), Azure OpenAI (22), Mistral AI (15). Each alternative, with the situation it fits, is on the DeepSeek alternatives page.

How much does DeepSeek cost per month?

At 10,000 users, DeepSeek costs $27 a month on the cheapest plan that fits (deepseek-flash), from its pricing page.

Do developers switch to or away from DeepSeek?

On GitHub, 68 public pull requests moved projects to DeepSeek from tools we list, most from Gemini API (24), OpenAI API (21), Claude (16); 48 moved them away, most to OpenAI API (16), Gemini API (15), Groq (9).

When should you use DeepSeek?

Token cost dominates your budget, for example high-volume batch or agent workloads (LLM API). Something else fits better if: you want the strongest general models and the widest ecosystem of examples and integrations, pick OpenAI API or Claude or Gemini API; you want to try or mix many models with one key and one bill, and fall back when a provider is down, pick OpenRouter.

Where is my data processed?

DeepSeek's privacy policy says it collects, processes and stores personal data in the People's Republic of China. The policy also says that for apps built on the Open Platform, the developer is the controller of its end users' data and must disclose its own policy to them. source ↗

Can I use the OpenAI or Anthropic SDK?

Yes. Set the base URL to https://api.deepseek.com for the OpenAI SDK or https://api.deepseek.com/anthropic for the Anthropic SDK, and use a DeepSeek API key. source ↗

What are the rate limits?

Each account has a cap on concurrent requests per model, whichever key is used, and exceeding it returns HTTP 429. You can ask for higher concurrency, which DeepSeek grants at no extra cost based on your needs. source ↗

Do I need to set up prompt caching?

No. Disk-based context caching is enabled for all users by default, and unused cache entries are cleared automatically, usually within a few hours to a few days. source ↗

Can I turn off the reasoning step?

Yes. Thinking mode is on by default at high effort; you can disable it or set the effort with the thinking and reasoning_effort parameters (or their Anthropic and Responses equivalents). source ↗

Who owns the outputs, and can I train on them?

You keep your rights in the inputs and DeepSeek assigns its rights in the outputs to you. The terms explicitly allow using inputs and outputs to train other models, including distillation. source ↗

Can I run the models on my own infrastructure?

The V4 models are released as open weights on Hugging Face, alongside a technical report, so you can serve them yourself or through another host. source ↗

Is DeepSeek free?

No permanent free tier; paid use starts at Pay per token. source ↗

Is DeepSeek open source or self-hostable?

Not open source, and hosted only.

Can AI coding agents work with DeepSeek?

No llms.txt, official MCP server or CLI found yet.

Who uses DeepSeek?

11 makers' products we track, each with a source, and 25 open-source projects declare it in their code. The best known of them on Product Hunt: MGX (Now Atoms) (3,165 followers), Genstore.ai (1,812 followers), Blink.new (1,524 followers), IFTTT (703 followers), Webhound (692 followers). source ↗