Subscriptions and builders
The lowest step of the market: subscriptions to neural networks for staff, and chatbots from builders. A builder chatbot costs from $6: it answers customers by a script with buttons, seldom writes to the CRM or the accounting system, and you are responsible for the result.
A platform agent on a subscription answers customers and can write to the CRM if you set it up yourself. The price follows the platform’s plan, and the settings, scripts and dialog history stay on its side for as long as you pay.
This step fits when you need to answer “are you open?” and “where are you located?”. If a request must turn into a deal in the CRM on its own, and the price list must come from the accounting system, a builder is not enough.
The hidden part of the subscription price: employees’ personal accounts without a contract, where customer details end up. This risk is covered on the page about leaks through neural networks.
Turnkey AI agents
A turnkey agent is built into the CRM, the accounting system and people’s work, and the contractor is responsible for the result under the contract. Our prices are fixed in the contract before the start, and the success metric is written there too:
1. Pilot: one AI agent on one process, from $1,499, 14 days.
2. AI department: 3–5 AI agents on shared data, from $4,999, 4–6 weeks. Integration with the CRM, accounting system and phone system is included in the price.
3. Turnkey business: several departments and a management dashboard, from $18,000, 2–3 months.
4. Support: from $480 a month, and you can stop in any month.
The difference from a subscription is not only the price. After the rollout you keep the system, the license, the data and the access rights, not the right to use an agent while you pay.
There is a blog article with a review of our plans and what stands behind them, and the full list is on the pricing page.
Development
Custom development is needed when the market has no ready-made solution: a rare process, unusual systems, your own storage requirements. Its price is based on the scope of work, so you can compare offers only for the same specification.
The model’s work adds to the development price. API models are paid per token: the professional model of a large Russian bank costs $6 per million tokens (16 September 2026), and the professional model of a Russian search company $14 per million tokens (13 September 2026).
So look at two numbers: the one-time development price and the monthly bill for the model and support. An offer without the second number shows only half the costs.
Before you pay for development, check whether a ready-made service or agent exists for the same process. A contractor often offers what it can build itself. How to tell when a ready-made solution is enough is covered separately.
Local models
A local model runs on your hardware or on rented graphics cards in Russia, and the correspondence does not go to an outside model provider. You pay for hardware, electricity, hosting and an engineer, not for tokens.
Prices with check dates:
1. RTX 4090 rental in Russia: from about $0.40 an hour (30 September 2026).
2. H100 rental: from $3 an hour (30 September 2026).
3. Mac Studio M4 Max with 128 GB of memory: $11,000 (29 September 2026).
4. H100 80 GB graphics card: $28,000 (29 September 2026).
Rental fits when you want to test a model’s quality on your own dialogs before you buy hardware. A purchase makes sense when the correspondence must not leave the company, and the volume of requests is large and stable.
Memory and total cost calculators, a comparison with the cloud and prices of ready-made configurations are collected in the section on the cost of sovereign AI.
How to compare prices
Two prices are comparable only if the same result stands behind them. Before you compare, bring the offers to one line:
1. What is included: only answers in a chat, or entries in the CRM, accounting system and phone system.
2. Who is responsible for the result: you, or the contractor by a metric in the contract.
3. What a month after launch costs: the model, support, edits.
4. Who owns the system after the rollout: the rules, the data, the access rights.
5. Where customer data is processed: in Russia or abroad.
The different answers to these questions usually explain the difference in price. A bot from $6 and a turnkey agent solve different tasks, and you choose between them by the process, not by the price tag.
Payback is calculated on your own numbers: how many employees do routine work, what they cost with taxes, and what share of the work AI will take. The calculator on the home page does this with your inputs, and in a 30-minute review we check the calculation with you.