Pricing
The licence buys the platform. The usage buys the work it does on your behalf. Both are priced below.
ISO/IEC 27001:2022 certifiedLicences
One shared memory, the assistant on top of it, and the connections into the systems you already run — Gmail and Outlook, Drive and OneDrive, Calendar and Teams, and more than a thousand others. Connecting a suite is one consent rather than an application at a time.
Everything that is organisational rather than personal — governance, the security and procurement work, agreed service levels, and a named route to support. One spending ceiling for the whole organisation, one access model across every department, and terms your legal team can sign.
In addition to the per-person licences · where it starts, not where it lands
Usage is billed separately either way.
Half an hour on your own numbers. You leave with the per-person figure and the organisation-wide one, and nothing to install to get them.
Usage
Usage is the AI work itself — the thinking, the listening, the remembering, the making — charged on top of the licence, and only for what you actually use. Every rate below is read from the same catalogue the product bills against, so a rate that changes there changes here.
Meter 01
A Synapse is the unit of AI work — one meter across the whole platform, whichever model does the work. Every rate below is per million Synapses.
Synapses are not tokens. A rate published elsewhere as a price per million tokens is a different quantity from the figures below, so the two do not compare directly. Every rate here is also restated in tokens, per model, in the machine-readable pricing document.
Spending limits apply per organisation, per API key and per person, over a day, a week or a month, with alerts on thresholds you choose.
Cached input is the rate for context the platform has already read once — long documents and long conversations spend most of their input there.
| Model | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| FTG 4 Pro Preview | CHF 3.30 | CHF 0.33 | CHF 16.50 |
| FTG 4 (Preview) | CHF 1.92 | CHF 0.15 | CHF 6.00 |
| FTG 4.0 Speed | CHF 0.70 | CHF 0.07 | CHF 1.20 |
| FTG Ultra | CHF 2.64 | CHF 0.264 | CHF 13.20 |
| FTG 3.0 Pro | CHF 13.20 | CHF 1.32 | CHF 66.00 |
| FTG 3.0 | CHF 2.64 | CHF 0.264 | CHF 13.20 |
Pick any of these instead and they meter identically, on the same invoice. Rates are the provider’s and move when they move them.
| Model | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| Claude Opus 5 | CHF 4.40 | CHF 0.44 | CHF 22.00 |
| Claude Sonnet 5 | CHF 1.76 | CHF 0.176 | CHF 8.80 |
| Claude Fable 5 | CHF 8.80 | CHF 0.88 | CHF 44.00 |
| Claude Opus 4.8 | CHF 4.40 | CHF 0.44 | CHF 22.00 |
| Claude Haiku 4.5 | CHF 0.88 | CHF 0.088 | CHF 4.40 |
| Model | Input / 1M | Output / 1M |
|---|---|---|
| GPT-OSS 120B (Turbo) | CHF 0.308 | CHF 0.66 |
| Model | Input / 1M | Output / 1M |
|---|---|---|
| Cohere Command A (03-2025) | CHF 2.20 | CHF 8.80 |
| Command A Vision | CHF 2.20 | CHF 8.80 |
| Command A Reasoning | CHF 2.20 | CHF 8.80 |
| Model | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| GPT-OSS 20B (Fast) | CHF 0.0616 | CHF 0.0308 | CHF 0.264 |
| Inkling | CHF 0.88 | CHF 0.1496 | CHF 3.564 |
| Kimi K2.7 Code | CHF 0.836 | CHF 0.1672 | CHF 3.52 |
| MiniMax M2.7 | CHF 0.264 | CHF 0.0528 | CHF 1.056 |
| MiniMax M3 | CHF 0.264 | CHF 0.0528 | CHF 1.056 |
| Muse Glimmer 30B | CHF 0.308 | CHF 0.0352 | CHF 1.32 |
| Nemotron 3 Ultra (Preview) | CHF 0.528 | CHF 0.1056 | CHF 2.112 |
| Nemotron 3.5 Lightning 30B | CHF 0.044 | CHF 0.0088 | CHF 0.176 |
| Qwen 3.7 Plus | CHF 0.352 | CHF 0.0704 | CHF 1.408 |
| Qwen 3.8 2.4T | CHF 1.76 | CHF 0.22 | CHF 5.28 |
| Qwen 3.8 Max | CHF 1.76 | CHF 0.22 | CHF 5.28 |
| Kimi K3 | CHF 2.64 | CHF 0.264 | CHF 13.20 |
| GPT-OSS 120B (Fast) | CHF 0.132 | CHF 0.0132 | CHF 0.528 |
| DeepSeek V4 Pro | CHF 1.5312 | CHF 0.1276 | CHF 3.0624 |
| GLM 5.2 | CHF 1.232 | CHF 0.1232 | CHF 3.872 |
| Kimi K2.6 (Vision) | CHF 0.836 | CHF 0.1408 | CHF 3.52 |
| Model | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| Gemini 3.5 Flash Lite | CHF 0.264 | CHF 0.0264 | CHF 2.20 |
| Gemini 3.6 Flash | CHF 0.66 | CHF 0.066 | CHF 3.30 |
| Gemini 3.7 Flash | CHF 0.66 | CHF 0.066 | CHF 3.30 |
| Gemini 3.1 Flash Lite Image | CHF 0.20 | CHF 0.05 | CHF 0.80 |
| Gemini 3 Pro Image | CHF 1.76 | — | CHF 10.56 |
| Gemini 3.5 Flash | CHF 1.32 | CHF 0.33 | CHF 7.92 |
| Gemma 4 31B | CHF 0.0528 | CHF 0.0132 | CHF 0.308 |
| Gemini 3.1 Pro (Preview) | CHF 1.76 | CHF 0.44 | CHF 10.56 |
| Gemini 3.1 Pro Custom Tools | CHF 5.00 | CHF 1.25 | CHF 15.00 |
| Gemini 3.1 Flash Lite | CHF 0.22 | CHF 0.055 | CHF 1.32 |
| Model | Input / 1M | Output / 1M |
|---|---|---|
| Groq Compound | CHF 0.5192 | CHF 0.6952 |
| Groq Compound Mini | CHF 0.5192 | CHF 0.6952 |
| Model | Input / 1M | Output / 1M |
|---|---|---|
| Mistral Medium 3.5 | CHF 1.32 | CHF 6.60 |
| Mistral Large | CHF 1.76 | CHF 5.28 |
| Codestral | CHF 0.264 | CHF 0.792 |
| Magistral Medium | CHF 1.76 | CHF 4.40 |
| Model | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| GPT-5.6 Sol | CHF 4.40 | CHF 1.10 | CHF 26.40 |
| GPT-5.6 Luna | CHF 0.88 | CHF 0.22 | CHF 5.28 |
| GPT-5.6 Terra | CHF 2.20 | CHF 0.55 | CHF 13.20 |
| GPT-5.5 Pro | CHF 13.20 | CHF 3.30 | CHF 79.20 |
| GPT-5.4 Mini | CHF 0.66 | CHF 0.165 | CHF 3.96 |
| GPT-5.4 Nano | CHF 0.176 | CHF 0.044 | CHF 1.10 |
Meter 02
Transcription is priced per hour of audio. The extra passes — who spoke, what it was about, what was decided — are priced individually on top and are off by default.
| Transcription | Rate |
|---|---|
| Universal-Streaming Multilingual | CHF 0.30 / hour |
| Add-ons, off by default | |
| Speaker Identification | CHF 0.05 / hour |
| Custom Formatting | CHF 0.08 / hour |
| Entity Detection | CHF 0.08 / hour |
| Sentiment Analysis | CHF 0.05 / hour |
| Auto Chapters | CHF 0.16 / hour |
| Key Phrases | CHF 0.03 / hour |
| Topic Detection | CHF 0.38 / hour |
| Summarization | CHF 0.08 / hour |
Meter 03
What the company remembers is held as units called Neurons — one Neuron is 10KB of indexed, searchable knowledge. Synapses are the work of putting something into that memory; Neurons are what holding it costs afterwards.
| Knowledge held | Price |
|---|---|
| 1,000 Neurons — about 10 MB | CHF 0.008 |
| 100,000 Neurons — about 1 GB | CHF 0.80 |
| 1,000,000 Neurons — about 10 GB | CHF 8.00 |
Meter 04
Images are priced per image rather than per Synapse, at a rate that depends on which model makes it. Upscaling an image you already have meters the same way.
| Model | Rate |
|---|---|
| ClipDrop Image Upscale | CHF 0.40 / image |
| DALL·E 3 | CHF 0.16 / image |
| DALL·E 2 | CHF 0.08 / image |
| Gemini 3 Flash Image Preview | CHF 0.15 / image |
| Gemini 3 Pro Image Preview | CHF 0.30 / image |
Image rates are the 2026-08-12 catalogue. They are the one set on this page the live rate feed does not carry yet.
One meeting touches three meters: the hour of audio, the Synapses for summarising it and filing what matters, and the Neurons that memory then occupies.
Estimate
Nothing here is sent anywhere — it runs in your browser against the rates above.
The arithmetic
CHF 30 a month is about CHF 1.50 per working day — roughly ten minutes of a mid-salary hour.
Take one person and estimate the time they lose in a normal week to looking for something that already exists, rebuilding something that was already built, or being the bridge between two systems.
Above an hour a week and the licence is already paid for. Below it, you do not need this yet — and it is better for both of us that you find that out during the trial rather than in month four.
What a pile of per-seat AI subscriptions costs a company with no shared memory, and what replaces them.
Services
Optional, and genuinely optional — most teams never buy any of it. When you do, it is our own engineers.
Where AI earns its place, and where it does not — including work that never touches our platform. Then how your people actually use it.
Connectors to systems only you run. Skills and flows for the work only you do. Features built into the platform when what you need is not there.
The interfaces and workflows your people actually meet, laid out so the first week does not need a manual.
Priced per engagement
Before the first invoice
Nothing continues by accident. Nothing charges unless you choose to carry on, and what you put in during the seven days is deleted unless you carry on. If you do carry on, it converts to Starter at CHF 30 per person per month.
The unit of AI work, across the whole platform. Every time Aitronos thinks on your behalf it spends Synapses — a question answered, a meeting summarised, a document read and filed into the company memory — whichever model does the work. One meter for all of it, and the rate for every model is published on this page.
Only if you switch the limits off. Spending ceilings apply per organisation, per API key and per person, and usage arrives broken down per person and per department rather than as one number at the end of the month — so an unusual week has a name and a reason attached to it before the invoice does.
No. The licence is per person, so the bill follows the number of people who actually hold one. Start with the team that will genuinely use Aitronos and add the rest when they ask for it.
Input is what you send — the question, the documents, the conversation so far. Output is what gets written back. Generating text costs materially more than reading it, and the rates say so rather than averaging the difference into one number. Context the platform has already read once meters at the cached rate, which is lower again.
The EU and Switzerland, contractually, on a platform certified to ISO/IEC 27001:2022 — certificate IC-IS-2606004, issued by InterCert, and available from the Trust Center at trust.aitronos.com. Running Aitronos on your own infrastructure is arranged rather than selected.
Start here
Bring your headcount and the work you would put through it first, and the question stops being theoretical.
Buying for the whole organisation? Bring that to the demo — half an hour on your own headcount, and you leave holding the organisation-wide number as well as the per-person one.
Questions? support@aitronos.com
