Let's get started,
one step at a time

Before you begin

This part is written for people with no development experience. If some terms are unfamiliar, that is perfectly fine — we explain each one as it comes up. Just read from the top and work your way down.

0.1Which models you can use

The AI Initiative covers four things: Solar Pro 2, Solar Pro 3, Solar Pro 4, and Document Parse. Choose from within these four.

Pro 2, Pro 3, or Pro 4?

If you are starting now, use Solar Pro 4. All three are covered by this program, and they share the same address and the same rate limits. When the terms are identical, take the newest generation. The official docs and the cookbook are also written around solar-pro4, so you can follow them as they are.

The three models take the same request format. To move between them, change the model value and leave everything else alone.

0.2Four things to prepare

An account, a payment method, an API key, and a place to run code. Once these four are ready, you can make your first call. Let's go through them one at a time.

0.3Understanding rate limits

AI Initiative accounts are provided at Tier 0. There is a limit on how many requests and how much text you can send per minute. Knowing this in advance helps when you plan a class or a large batch.

The same limits apply to Solar Pro 2, Solar Pro 3, and Solar Pro 4. You can check your current tier and the limits that come with it under Billing → Commitment → Rate limit in the console.

Why this matters for this program in particular

AI Initiative applicants work at universities, university hospitals, and nonprofits. Situations where many requests arrive at once are far more common here than one person calling the API occasionally.

If you receive a 429

A 429 is not a failure — it means "please slow down for a moment." Here is what to do.

1. Do not resend immediately. Wait a few seconds and try again. If it repeats, increase the wait gradually.
2. Reduce how many requests you run at the same time. If you were sending 10 at once, try 2 or 3.
3. In a classroom, split into groups and run in turns.

Checking how much you have left

Every response includes your remaining allowance. The command below shows only that information instead of the response body.

curl -s -D - -o /dev/null \
  -H "Authorization: Bearer $UPSTAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "solar-pro4", "messages": [{"role": "user", "content": "Hello"}]}' \
  https://api.upstage.ai/v1/chat/completions \
  | grep -i "X-Upstage-RateLimit"

0.4Your first response

Now let's actually run something. The goal of this section is one successful call each to Solar and Document Parse.

① Say hello to Solar

Paste the code below into your terminal and run it. A terminal is the window where you type commands directly — PowerShell on Windows, the Terminal app on macOS.

Replace $UPSTAGE_API_KEY with the key you created earlier.

curl -X POST https://api.upstage.ai/v1/chat/completions \
  -H "Authorization: Bearer $UPSTAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "solar-pro4",
    "messages": [
      {"role": "user", "content": "Introduce yourself in one sentence."}
    ]
  }'

After a moment you will get back a long response wrapped in braces. It looks complex at first, but there are only a few parts you actually need.

With Python the setup is even simpler. Solar is compatible with the OpenAI API, so you can use the familiar OpenAI library and only change the address.

pip install --upgrade openai

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["UPSTAGE_API_KEY"],
    base_url="https://api.upstage.ai/v1",
)

response = client.chat.completions.create(
    model="solar-pro4",
    messages=[{"role": "user", "content": "Introduce yourself in one sentence."}],
)

print(response.choices[0].message.content)

② Turn a document into text

Now for Document Parse, which converts a PDF or image into a form a computer can read. The one thing to remember is that it uses a different address from Solar.

Put a PDF named sample.pdf in the same folder and run this.

curl -X POST https://api.upstage.ai/v1/document-digitization \
  -H "Authorization: Bearer $UPSTAGE_API_KEY" \
  -F "document=@sample.pdf" \
  -F "model=document-parse" \
  -F "ocr=force"

You have made it this far

• You saw a sentence written by the model in the Solar response.
• You saw a response containing your document's contents from Document Parse.
• Both requests completed without errors.

If all three are true, the hands-on part of Part 0 is complete. Nicely done.

0.5[Activity] Setup checklist

Copy the template below and fill it in yourself. Writing it out once makes the following parts much easier to follow.

# AI Initiative setup checklist

## 1. Account and credits
- [ ] Signed in at console.upstage.ai
- Email used:
- [ ] Checked balance under Billing → Credit
- Current balance:

## 2. Payment method
- [ ] Registered a payment method under Billing

## 3. API key
- [ ] Created a key starting with up_
- [ ] Stored the key somewhere safe
  (In Colab — secret name: UPSTAGE_API_KEY / notebook access ON)

## 4. First response
- [ ] Got a Solar response — model used:
- [ ] Got a Document Parse response
- If there was an error, the status code:

## 5. My plan
- Organization type (university / university hospital / nonprofit or NGO):
- Work I want to automate:
- Expected scale (how many at a time, how many people at once):
  → Does this fit within the Tier 0 limits in 0.3?

Checkpoint

• All five items are filled in.
• You have seen a response from Solar and from Document Parse at least once each.
• You have judged whether your expected scale fits within the Tier 0 limits.
• If it does not, you have written one line about how you will split the requests.

0.6Frequently asked questions

The six things people ask most often when starting out. Find your symptom below.

Where to go next

Now that the setup is done, let's look at what these tools actually do.