How to Get an LLM: 4 Practical Ways to Start

"How to get an LLM" usually means one of two things: you want to use a large language model through a chat interface for everyday tasks, or you want to integrate one into an application via code. The right path depends entirely on which of those two you're after — and the setup time ranges from thirty seconds to several days.

This guide covers the four realistic ways to get access to an LLM in 2025: consumer chat apps, official provider APIs, self-hosted open-weight models, and hosted API gateways that sit on top of an existing subscription. Each has different costs, setup effort, and tradeoffs.

Option 1: Use a Chat App (Fastest, No Code)

If you just want an LLM to answer questions, write drafts, or help you think through a problem, sign up for a consumer product:

• Claude.ai (Anthropic) — strong reasoning, long context, good for writing and analysis
• ChatGPT (OpenAI) — broadest plugin/tool ecosystem
• Gemini (Google) — tight integration with Google Workspace

Setup is: go to the site, create an account, start typing. Free tiers exist but are rate-limited; paid plans run roughly $20/month for individual use. This is the right choice if you never need to call the model from your own code.

Option 2: Use an Official Developer API

If you're building a product — a chatbot, a content pipeline, an internal tool — you need programmatic access, not a chat window. Each major lab offers a developer API:

1. Create a developer account (separate from your consumer subscription in most cases)
2. Add a payment method and get billed per token
3. Generate an API key
4. Send HTTP requests to their endpoint

A minimal request looks like this:

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Explain quicksort in one paragraph"}]
  }'

This is the standard path for most developers. Downsides: you're billed per token with no monthly cap, pricing tiers and rate limits differ from your personal chat subscription, and if you're already paying for a consumer plan, you're now paying twice — once for chat, once for API usage.

Option 3: Self-Host an Open-Weight Model

If you need full control — data never leaves your infrastructure, no per-token billing, custom fine-tuning — you can run an open-weight model yourself.

Common choices: Llama 3, Mistral, Qwen, DeepSeek. Typical setup:

# Using Ollama as an example
curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3
ollama run llama3

For production use you'll want a proper inference server (vLLM, TGI, or similar) and a GPU with enough VRAM — a 70B model needs roughly 140GB+ for full precision, less if quantized. This is the right option when data residency or cost-at-scale matters more than engineering time. It's the wrong option if you want something running today: model selection, quantization, prompt formatting, and hardware provisioning all take real effort, and open-weight models generally lag behind frontier closed models on complex reasoning tasks.

Option 4: Turn an Existing Subscription Into an API

There's a fourth path that's often overlooked: if you already pay for a Claude subscription for personal or team use, you don't necessarily need a separate developer account with its own billing. SubToAPI turns your existing Claude access into a standard HTTPS API — you get an application key (sub_live_...), send normal /v1/messages requests, and usage draws from the plan you're already paying for instead of a second per-token bill.

This matters if you're prototyping, building an internal tool, or shipping a small feature and don't want to stand up separate API billing, negotiate rate limits, or manage two overlapping subscriptions.

const response = await fetch("https://api.subtoapi.app/v1/messages", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.SUBTOAPI_KEY}`,
    "content-type": "application/json"
  },
  body: JSON.stringify({
    model: "claude-sonnet-4-5",
    max_tokens: 1024,
    messages: [{ role: "user", content: "Summarize this changelog in 3 bullets" }]
  })
});

const data = await response.json();
console.log(data.content);

The API supports streaming, tool use, and returns usage metadata per request, which matters if you're tracking cost per feature or per customer. Setup takes a few minutes: sign up, generate a key, start calling the API. See the quickstart for the full flow, or the messages and streaming docs for request formats.

Which Option Should You Pick?

• Just want an assistant? Use a chat app.
• Building a product, already have a Claude subscription? Start with SubToAPI — check pricing for Solo (€9), Team (€19/seat), and Scale (€49/seat) plans.
• Building a product, no existing subscription? Go direct to the provider's developer API.
• Need full data control or heavy custom fine-tuning? Self-host an open-weight model.

Most teams end up combining two of these: a chat app for individual work, and an API — official or via a gateway like SubToAPI — for anything shipped into a product.

Questions

Is there a free way to get an LLM? Yes. Most chat apps (Claude.ai, ChatGPT, Gemini) offer free tiers with rate limits. Self-hosted open-weight models are also free to run if you already have the hardware, though inference costs (electricity, GPU time) aren't zero.

Do I need a developer account to call an LLM from code? Not necessarily. Official provider APIs require their own developer account and billing, but tools like SubToAPI let you call the API using an existing Claude subscription instead of setting up separate API billing.

What's the difference between an LLM API and a chat app? A chat app is a finished product with a UI, built for humans typing questions. An API is a programmatic interface — you send structured requests (like JSON with a messages array) and get structured responses back, meant for integrating into your own software.