One key. Big open-weight AI models. Credits that keep piling up.
Virsilipas is our OpenAI-compatible AI API. Point any tool that takes a base URL at it and use GLM-5.3, DeepSeek-V4-Pro and eight more open-weight models on one key, paid in credits that currently never expire.
Get credits from 9.99€ a month See it running live
Live numbers: requests, tokens and success rate of the last 7 days are on virsilipas.mcx.fi.
What you get
The big models
GLM-5.3 and DeepSeek-V4-Pro with a context of a million tokens, plus thinking and instant variants. Fast models for the everyday work, and models with few refusals when you need them.
Works with your tools
OpenAI-compatible chat completions at https://virsilipas.mcx.fi/v1. The same key also speaks the Anthropic Messages format (Claude Code) and the Gemini API format (Gemini CLI).
Credits that stay
Credits currently never expire. What you do not use this month is still there next month, and every renewal adds more on top.
Virsilipas AI Credit Plans
- Virsilipas 100
- 9.99€
monthly billing - 9,990,000 credits every month
- Every euro you pay comes back as credits
- Up to 6.2 million input tokens on GLM-5.3
- Or up to 37.3 million on the fast default model
- Credits on your key within minutes of payment
- Currently never expire
- Virsilipas 125
- 49.99€
monthly billing - 62,487,500 credits every month
- 25 % more credits per euro
- Up to 38.9 million input tokens on GLM-5.3
- Or up to 233 million on the fast default model
- Credits on your key within minutes of payment
- Currently never expire
- Virsilipas 150
- 299.99€
monthly billing - 449,985,000 credits every month
- 50 % more credits per euro
- Up to 280 million input tokens on GLM-5.3
- Or up to 1.7 billion on the fast default model
- Credits on your key within minutes of payment
- Currently never expire
We may change the plans and their credit amounts at any time.
Months on the plan, credits left if none are spent. Under the current rules nothing expires; spend them whenever you like.
Need more? Order a plan several times: each one adds its own credits every month, all to the same key. Prices exclude VAT, like all our listed prices. A refund takes back the credits that payment added.
The models
Ask for a model by its alias. The two largest open-weight models lead the list; every alias below is on the same key and the same balance.
GLM-5.3 · large
The strongest model on the key: one million tokens of context, thinks before it answers. large-think thinks at length, large-instant answers straight away.
1.61€ per million input tokens, 7.15€ per million output tokens on the 9.99€ plan.
DeepSeek-V4-Pro · alt-large
A second large model with a different style: one million tokens of context, answers directly, and alt-large-think reasons first.
2.32€ per million input tokens, 4.64€ per million output tokens on the 9.99€ plan.
| Alias | Model | Context | € / M input | € / M cached input | € / M output | Good for |
|---|---|---|---|---|---|---|
large | GLM-5.3 | 1M | 1.61 | 0.36 | 7.15 | Large, thinks before it answers |
alt-large | DeepSeek-V4-Pro | 1M | 2.32 | 0.18 | 4.64 | Large, answers directly; -think to reason |
nolimits-large | Hermes-3-Llama-3.1-405B | 128k | 1.79 | 1.79 | 1.79 | Large, few refusals |
default | GLM-5.3-Flash | 1M | 0.27 | 0.05 | 0.89 | Fast everyday model, the cheapest |
alt | DeepSeek-V4.1-Flash | 1M | 0.36 | 0.01 | 1.07 | Fast, very cheap on repeated input |
nolimits | Hermes-3-Llama-3.1-70B | 128k | 1.25 | 1.25 | 1.25 | Few refusals, fast |
nolimits-story | L3.1-70B-Euryale-v2.2 | 128k | 1.52 | 1.52 | 1.52 | Fiction and role-play |
nolimits-flash | GLM-5.3-Flash uncensored | 1M | 0.36 | 0.13 | 1.43 | Few refusals, newest generation |
nolimits-qwen | Qwen3.8-27B uncensored | 1M | 1.79 | 0.39 | 5.36 | Few refusals, mid-size |
nolimits-mimo | MiMo-V2.6-Flash | 1M | 0.89 | 0.49 | 2.68 | Few refusals, fast reasoning |
Euro prices are on the 9.99€ plan, where 1,000,000 credits cost 1€. The 49.99€ plan gives 1.25 times and the 299.99€ plan 1.5 times the credits per euro, so the same tokens cost a fifth and a third less there. Prices as of 18 October 2026; the live prices, aliases and descriptions are on the model cards.
Start in three steps
1 Pick a plan above and pay the invoice. Your credits land on your key within minutes, and again every month the plan renews.
2 Open your key: Client Area, then AI credits, then Virsilipas AI key. Your account has one key for everything, made for you on the first credits; it starts with vsk-. The same page shows your balance, your use and the credit history.
3 Point your tool at it: base URL https://virsilipas.mcx.fi/v1, your key, and a model alias such as large.
From a shell
curl https://virsilipas.mcx.fi/v1/chat/completions \
-H "Authorization: Bearer $KEY" \
-H 'Content-Type: application/json' \
-d '{"model": "large", "messages":
[{"role": "user",
"content": "Say hello."}]}'
curl https://virsilipas.mcx.fi/v1/balance \
-H "Authorization: Bearer $KEY"
From JavaScript (Node 18, no packages)
const KEY = process.env.VIRSILIPAS_KEY;
const URL = 'https://virsilipas.mcx.fi/v1';
const res = await fetch(`${URL}/chat/completions`, {
method: 'POST',
headers: { 'Authorization': `Bearer ${KEY}`,
'Content-Type': 'application/json' },
body: JSON.stringify({ model: 'alt-large',
messages: [{ role: 'user',
content: 'Say hello.' }] }),
});
const reply = await res.json();
console.log(reply.choices[0].message.content);
console.log(`credits: ${reply.usage.credits}`);
Claude Code: set ANTHROPIC_BASE_URL=https://virsilipas.mcx.fi and your key as ANTHROPIC_AUTH_TOKEN. Gemini CLI and every other setting are in the README on GitHub. On a Pulsed Media seedbox with a current PMSS, Codex, Claude Code and Gemini CLI start on your credits by themselves when ~/.virsilipasKey holds your key: see AI coding tools on PMSS.
Why Virsilipas
Your credits keep adding up. Under the current rules credits never expire. Buy a plan for a busy month and the leftovers wait for the next project.
Open source, out in the open. The credit service is mcxVirsilipas on GitHub. Its code never stores or logs the text of your prompts or answers; per request it records counts such as tokens and credits, the model, status and timing. Your prompt does go to the model service that answers it.
Watch it run. Requests, tokens, success rate and the model mix are public, hour by hour, on virsilipas.mcx.fi.
Already a customer? Your services already earn free credits on every invoice you pay, on this same key: see AI credits. A plan adds to that.
Pick your planQuestions
What is Virsilipas?
An OpenAI-compatible API for open-weight AI models, run by Pulsed Media. One key, one balance, paid in credits. The name is the song-chest of the Kalevala, opened before the singing starts.
Do the credits expire?
Currently no. Credits stay on your key and keep adding up month after month.
How fast do the credits arrive?
Within minutes of a paid invoice, for the first payment and for every renewal.
Can I buy more than one plan?
Yes. Order a plan as many times as you like; each adds its own credits every month to the same key.
What happens if I cancel?
The plan stops renewing and stops adding credits. Credits already on your key stay there. A refund takes back the credits that payment added.
Which tools work with it?
Anything that lets you set an OpenAI-compatible base URL and an API key, plus Claude Code through the Anthropic Messages format and Gemini CLI through the Gemini API format. Requests are text only.
Where do I see my balance?
On the Virsilipas AI key page in the Client Area, or with GET /v1/balance and GET /v1/usage on your key.