SIMUN
Rules Of Procedure Sample Position Paper Draft Resolution Template Sample Speeches How to Get an API Key Understanding API How it Works What is MUN? FAQ ContactSIMUN runs on a bring-your-own-key system. Instead of us charging you for every delegate and chair speech the AI generates on our server, you use your own API key and pay the AI company directly. There's no middleman markup or hidden charges, just the real cost of the AI work your simulation uses, which isn't much.
You may have only used AI through a website or chatbot like Claude, Gemini, or ChatGPT, so an "API key" may sound uncomfortably technical. But getting one is so simple. The videos on the left guide you through it in a few easy steps.
An API key (Application Programming Interface) allows one software to talk to another. SIMUN supports several AI models of varying cost and sophistication. Once you provide SIMUN with your chosen API key, it uses this to to run the AI simulation. The AI model directly charges you for every turn you play on SIMUN.
Like a password, an API key is a string of letters and numbers sk-ant-xxxxxxxx that proves the request, and the bill for it, belongs to you. SIMUN uses your key only to run your simulation. It never stores or shares it. Like any other password, you too should never share it with anyone or post it publicly.
AI credits are deducted for each turn during the simulation, but the amount for each turn is tiny and there are many turns per MUN session. So the real question is, how much does it cost to play an entire MUN session right up to the conclusion where the best delegate is announced.
API billing works like a prepaid phone plan. You load a small amount of credit onto your key, and it ticks down as SIMUN uses the model on your behalf. Providers typically suggest starting with around $5 of credit, which at roughly $2 a session means you can comfortably run a committee two times over, try a different topic, or replay a session with a different approach.
For the price of a Starbucks coffee that vanishes in under thirty minutes, you can enjoy a full day of diplomatic intrigue.
Anthropic is the company behind Claude. Their developer platform is called the Claude Console.
sk-ant-. If you lose a key, you can't recover it, just delete it in the Console and create a new one.
Claude comes in three main tiers. All of them are paid but the price difference between tiers is large, so picking the right one matters a lot for your budget.
Cheapest and fastest. Fine for exploring the interface or a throwaway test, but too unspohisticated for a meaningful complete session.
The best value for money. Delegates hold consistent positions, respond meaningfully, and the chair enforces the ROP at a reasonable price of about $2 for the entire committee.
Capable, witha hefty price tag. Costs 2-3 times more than Sonner. Allows for the sparpest possible debate, but for most practice and learning purposes it's overkill.
A simple rule of thumb: use Haiku only to confirm your key is working, run Sonnet for realistic debate and only reach for Opus if you want the best quality available and have the budget for it.
Google's AI models are called Gemini. The only platfrom supported by SIMUN that provides a free key is Google AI Studio.
Cheapest tier. Fast and inexpensive; usable for a bare-bones test run, but delegates can feel a little flat.
Often overlaoded. Overused due to its free availability, and practically unsuable during most of the day.
Most capable, paid only. No free option and noticibly more expensive than Flash. Reserve it for high-stakes practice.
OpenAI makes the GPT models behind ChatGPT. Their developer site, the OpenAI Platform, is a completely separate system from ChatGPT.com — so a ChatGPT Plus subscription does not give you API access or credit, and API usage has to be paid for separately.
sk-proj-. Treat it like a password: never paste it into code you publish publicly or share in a chat.
OpenAI's current line-up follows a similar "small, medium, large" pattern to Claude's and Gemini's, just with different names.
Cheapest, fastest. Built for simple, high-volume, low-stakes tasks. Not sufficient for an immersive MUN experience.
A solid middle ground. Noticeably better reasoning than Nano for a moderate price increase. Good for fun or low-stakes experiments.
The full-size flagship model (and its slower "thinking" variants, which take a while). Best quality and the best choice for a tough and realistic challenge.
OpenAI renames and re-releases these fairly often (you may see version numbers attached, like GPT-5 or newer), but the nano / mini / full-size pattern — and the price ladder that goes with it — tends to stay the same.
Side-by-side
Here's the short version, to help you decide where to start. "Session cost" is for a full SIMUN committee on each provider's recommended mid-tier model.
| Provider | Where to get a key | Free tier? | Full session cost |
|---|---|---|---|
| Anthropic (Claude) | console.anthropic.com | No — small starting credit required | ~$2 on Sonnet — our overall recommendation: reliable, realistic, and the best value per dollar |
| Google (Gemini) | aistudio.google.com | Technically yes, but frequently overloaded — not dependable for a full session | Lower per-token than Sonnet, not dependable |
| OpenAI (ChatGPT) | platform.openai.com | No — separate from any ChatGPT subscription | Varies — Mini is a reasonable mid-tier starting point |