Recommended1M context
Agent
The default for agent reasoning and tool calls.
Cached input
€0.05
per million tokens · €0.0595 incl. VAT
soloniq/agent
The European gateway for open-weight AI
Open models. Kept in Europe.
DeepSeek, GLM, Kimi, Qwen and MiniMax — served from servers in the EU. One API for every model. Nothing stored.
$ export OPENAI_BASE_URL=https://api.soloniq.com/v1
$ agent run --model soloniq/agent
› Customer asks: “Can you move my invoice to next month?”
› tool_call billing.reschedule(invoice: "INV-2291", to: "2026-11")
Done — INV-2291 now falls due on 1 November. I’ve emailed you the updated invoice.
Region: EUFallback model: readyCached input: 91%Prompt stored: never
The Soloniq Router
Every request, routed. Never stranded.
One address in front of EU-hosted models. The router checks your key, maps your model name to the model serving it today, and moves to the backup the moment the default falters.
Swipe to follow the route ›
Our names, not theirs.
You call soloniq/agent. We decide which EU-hosted model serves it today.
Fallback in milliseconds.
Errors or slow replies move traffic to the backup model. Your agent keeps working.
Budgets enforced.
Every key is checked against its budget and your balance before a token is spent.
Built to be swapped.
Accounts and keys live in our own layer, so the models behind a name can change without you noticing.
Sovereign by design.
Not a promise on a landing page. A rule enforced on every request.
Prompts and replies stored
Not by us, not by the providers we route to. We count tokens — never words.
Where your data goes
EU only.
Servers in the EU. Inference, monitoring, logs and backups too.
Verified
Model by model.
A model is sold only when it runs on servers in the EU, under a signed data processing agreement.
Training on your data
Never.
Zero retention and no training, written into every provider contract.
Billing
In euros.
EUR invoices, prepaid credits, a budget on every key.
The open-weight frontier
Frontier models. Open weights. European hands.
.env
OPENAI_BASE_URL=https://api.openai.com/v1
OPENAI_BASE_URL=https://api.soloniq.com/v1
MODEL=soloniq/agent
AUX_MODEL=soloniq/agent-fast
# same OpenAI SDK, same code — fallback is automatic
Every major open family.
DeepSeek, GLM, Kimi, Qwen and MiniMax behind one key. Switch models by changing one string — no new contracts, no new SDKs.
Tested on real workloads.
Each model is checked for tool calling, long context and prompt caching before it’s listed — so it behaves in production, not just on benchmarks.
Open weights, EU copies.
These models publish their weights. We serve copies running on servers in the EU, so your data never touches the original lab’s API.
Pick your model.
Alias1M context
Agent Fast
Summaries, routine steps and context compression.
Cached input
€0.05
per million tokens · €0.0595 incl. VAT
soloniq/agent-fast
Alibaba Qwen256K context
Qwen 3.8 Flash
Fast and cheap, sees images.
Cached input
€0.05
per million tokens · €0.0595 incl. VAT
soloniq/qwen-3.8-flash
MiniMax1M context
MiniMax M3
Long-context agents, sees images.
Cached input
€0.10
per million tokens · €0.119 incl. VAT
soloniq/minimax-m3
DeepSeek1M context
DeepSeek V4 Pro
Hard reasoning, maths and code.
Cached input
€0.50
per million tokens · €0.595 incl. VAT
soloniq/deepseek-v4-pro
Moonshot AI · Premium1M context
Kimi K3
Premium agent model with vision.
Cached input
€0.75
per million tokens · €0.8925 incl. VAT
soloniq/kimi-k3
10 models
All models and prices
GLM, DeepSeek, Kimi, Qwen and MiniMax — with what 1,000 agent turns cost on each.
See the price list ›
Agents resend their context constantly, so cached input is usually most of the bill — and the cheapest part.
The agent and agent-fast aliases run on GLM-5.3-Flash. If it is unavailable, requests go to DeepSeek V4.1 Flash and are billed at its price. Prices per million tokens in EUR, excluding VAT; private customers pay the price incl. 19% VAT shown on each card.
Made in Europe. Down to the rack.
Keys for teams
A key for every person, team and instance.
Give every developer, department, agent or server its own key — with your own ID, its own models and its own spending limits. One account, one invoice, full control.
Maria · Salesdept-sales · agent, agent-fast
€12.40 of €50Support agentagent-01 · agent, agent-pro
€124 of €200Server 7f3a · Customer 1042vps-7f3a · today €1.20 of €3
€21.50 of €25Research teamdept-research · kimi-k3
€40 of €500Your IDs, your names.
Tag each key with your own reference — an employee, a cost centre, a customer or a server.
Limits per key.
Monthly and daily budgets in euros, plus requests and tokens per minute. A key that hits its limit pauses — nothing else does.
Models per key.
The support agent gets agent-fast, the research team gets Kimi K3. Each key only reaches the models you allow.
See who spends what.
Usage per key and day, CSV export, and alerts at 80% and 100% of a budget. Partners automate all of it through the Partner API.
Security
A leaked key can only spend its own budget.
Agents never see keys.
Secrets are blanked in every tool result and outgoing message.
Untrusted stays untrusted.
Web, email and file content is labelled, with injection phrases flagged.
One key, one budget.
Stored hashed, shown once, rotated without downtime.
Stops on strange spend.
Per-key rate limits and daily caps stop runaway spending.
Partner programme
Your brand. Your billing. Our EU backbone.
Create a key per customer through the Partner API, set its budget in euros, and get usage back every night. Your customers never need a Soloniq account.
Your brand.
Your customers see your product and your invoice. Soloniq stays in the background.
A key per customer.
Create keys, set monthly and daily budgets in euros, and block or rotate them through the Partner API.
Usage every night.
Signed webhooks with each customer's usage, plus alerts at 80% and 100% of budget.
Your first key. In a minute.
Sign up with your email, add credit by PayPal or bank transfer, and create a key. Prepaid in euros, a budget on every key.