Hackathon London
Saturday 21 Nov 2026
Build with a European brain
A first call to a Mistral model in five minutes, with your own key or a model on your own laptop. Then build something a sovereign website could use.
Start here
- Pick a path. Your own Mistral key, or a local model with Ollama. No key needed for the local path.
- Make one call. Copy the five-minute recipe below.
- Open the demos. Run the same brief on Mistral beside another model.
- Build. Ideas below; anything that makes the model a choice counts.
- Share it. A public repository you own, and the link.
The three demos
- 1
Side by side
One brief to a European model, any model on OpenRouter and a model on your own computer, with the time taken and the tokens for each call.
Same brief, three brains - 2
Your own key
Where your key is kept (a variable in the tab, gone on reload) and where it goes (only to the provider you choose).
Your key, and where it goes - 3
On your device
Level 1: a model on your own machine, with nothing leaving it after the download.
On your device
Five minutes to a first call
A. With your own key
- Get an API key from the Mistral console. Free mode is the default for new accounts, with the lowest rate limits. Mistral’s key quickstart asks for “a plan with API access”; whether Free mode needs a payment card today was not confirmed. (Source: Mistral docs, Subscriptions and Mistral docs, Activate and generate an API key; as read on 20 November 2026.)
- In Free mode, Mistral may train on your API calls unless you switch it off in the Admin panel, under Privacy. Send nothing confidential. (Source: Mistral AI, Cloud inference and Mistral help centre, Do you use my data to train and Mistral help centre, Opting out of training; as read on 20 November 2026.)
- In a terminal, run
export MISTRAL_API_KEY=followed by your key. Never paste it into a shared document or a commit. - Run this:
curl https://api.mistral.ai/v1/chat/completions \
-H "Authorization: Bearer $MISTRAL_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "ministral-3b-2512",
"messages": [{"role": "user", "content": "Say hello in five words."}]}'
Ministral 3 3B, model name ministral-3b-2512, is listed as Apache 2.0, released 2 December 2025, at 0.10 dollars per million tokens in and 0.10 out. (Source: Mistral docs, Ministral 3 3B and Mistral docs, Pricing; as read on 20 November 2026.) For processing in the EU, change the address to api.eu.mistral.ai, at 1.1 times the price, once you have checked it serves the model. (Source: Mistral docs, Regional inference; as read on 20 November 2026.)
B. With a local model
- Install Ollama.
- Pull the smallest Ministral 3 and call it:
ollama pull ministral-3:3b
curl http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "ministral-3:3b",
"messages": [{"role": "user", "content": "Say hello in five words."}]}'
Ollama lists ministral-3 with the tags 3b, 8b and 14b. (Source: Ollama, ministral-3; as read on 20 November 2026.) Ollama serves an OpenAI-style endpoint at http://localhost:11434/v1/chat/completions. (Source: Ollama docs, OpenAI compatibility; as read on 20 November 2026.) Ministral 3 was announced on 2 December 2025 in 3B, 8B and 14B sizes, for edge and local use, each with image understanding, under Apache 2.0. (Source: Mistral AI, Mistral 3; as read on 20 November 2026.)
Same request, different address: that is the whole idea.
The model is a choice the owner makes, like the host.
What to build
- A model picker. One OpenAI-style interface, with the owner’s choice of model behind it.
- An open eval. The same brief on Mistral beside other models, with the tokens and time shown.
- A measurement. Ministral 3 on a Raspberry Pi or a tablet, timed and published. Today the fit is an estimate, not measured. (Source: A European brain for a sovereign website (Dinis Cruz, SGit Newsroom); as read on 20 November 2026.) Mistral’s official compressed (Q4_K_M) files are 2.15 GB for the 3B and 5.20 GB for the 8B, plus about 0.85 GB for vision. Those are download sizes, not memory use or speed. (Source: Hugging Face, Ministral-3-3B-Instruct-2512-GGUF and Hugging Face, Ministral-3-8B-Instruct-2512-GGUF; as read on 20 November 2026.)
- A sovereignty check. Who holds the keys, where it runs, which licence, how long data is kept, can you leave. See the ledger.
Before you serve a model to others, check its licence. Ministral 3, Mistral Small 4 and Voxtral Small are labelled Apache 2.0. Mistral Large 4 is under a modified MIT licence: its announcement says a company offering it as a model-as-a-service needs a separate commercial licence, while Mistral’s help centre describes the condition as a monthly revenue threshold of 20 million dollars. The pages disagree; read the licence file itself before relying on either. (Source: Mistral docs, Ministral 3 8B and Mistral AI, Mistral Large 4 and Mistral help centre, Licences of the open models; as read on 20 November 2026.)
How to share what you build
- Put it in a public repository you own, so it can move anywhere.
- Say in the README which path you used (your key, a local model, or another provider) and which model.
- Check there is no key in it before you push.
- Post the link where the organisers ask.
Venue, times, schedule and anything else about the day come from the organisers, not from this page.
The small print
This briefing follows Dinis Cruz’s article A European brain for a sovereign website (20 November 2026), which proposes working with Mistral. Nothing has been agreed with Mistral. Facts link their sources and were read on 20 November 2026; see all sources.