Agent and Knowledge Base Pricing

AI agents use large language models and a knowledge base. This section describes the available models, pricing, and how tokens are counted.

Available Models

We provide access to the following large language models:

• OpenAI GPT-5.1 (reasoning)
• OpenAI GPT-5.2 (reasoning)
• Claude 4.5 Haiku (reasoning)
• Claude 4.6 Opus (reasoning)
• Claude 4.6 Sonnet (reasoning)
• Gemini 3.1 Pro Preview (reasoning)
• Gemini 3 Flash Preview (reasoning)
• Gemini 2.5 Flash Lite
• Qwen 3 Flash
• Qwen 3 Max
• Qwen 3 Plus

How Tokens Work

Models process text by splitting it into smaller units called tokens. A token can be:

• part of a word (for example, "techno" + "logy");
• a short word;
• a symbol or punctuation mark.

On average, 1,000 tokens ≈ 750 words.

These values are approximate and provided only to give a general idea of usage costs. Each model uses its own tokenizer and splits text differently.

Tokens are billed for both input and output. For example, if your request contains 20 tokens and the model generates a response of 30 tokens, a total of 50 tokens will be deducted from your quota.

Reasoning Models

If a model displays a brain icon, it supports reasoning mode.

In this mode, the model generates intermediate reasoning steps before producing the final answer. It asks itself clarifying questions and answers them internally. This helps produce more accurate and well-structured responses, especially for complex tasks.

Each reasoning step requires additional input and output text, which increases token usage.

Reasoning mode cannot be disabled.

Agent Pricing

AI agents support two types of pricing:

• Base plans: fixed token packages that renew automatically every month.
• Additional token packages: one-time token purchases valid only until the end of the current billing period for the base plan.

Base Plans

Pricing for each model is available in the control panel when creating an agent.

The plan fee is charged once when the agent is created or when its plan is changed. In the following month, the plan renews with another one-time charge. Any unused tokens from the current month do not roll over.

The optimal number of tokens for your agent can only be determined during actual usage by monitoring consumption.

When getting started, we recommend selecting the smallest suitable plan and increasing it later if needed. Downgrading a plan is not supported.

Additional Token Packages

If you need more tokens but the next plan tier is too large, or if you only need additional tokens temporarily, you can purchase additional token packages in the control panel instead of changing your plan.

You can configure the number of tokens in increments of 250,000. The package remains valid only until the end of the current billing period (you can check the renewal date in Billing → Billing history).

The package fee is charged once. Your plan does not change, and the standard plan fee will be charged again next month.

Knowledge Base Pricing

How Price is Calculated

The cost of a knowledge base consists of two components:

• Token Subscription Fee. When a knowledge base is created, $5.63 is charged for 10 million tokens. These tokens are used for indexing the knowledge base and for all queries to it. This charge repeats monthly.

• Hourly Fee for the OpenSearch Database. The knowledge base runs on a cloud database. After the database is created, hourly billing starts. The cost depends on the selected configuration.

When creating the database, choose the smallest suitable configuration. You can increase it later if needed, but downgrading is not supported.

Pricing Example

When creating a knowledge base with the minimum configuration:

• $5.63 is charged immediately for 10 million tokens (monthly charge).
• $0.06831 per hour is charged for the database while it is running throughout the month.

Additional Tokens

If needed, you can add tokens at a price of $0.75 per 1,000,000 tokens by clicking Add Tokens on the knowledge base Dashboard:

Additional tokens are valid only until the end of the current billing period. For example, if the next billing date is in two days, the added tokens will expire in two days.

Token Usage During Indexing

Indexing uses the text-embedding-3-large embedding model, which converts text into vector representations. This is required for indexing data and enabling search and retrieval.

Token usage during indexing depends on the format of the uploaded file. Below is an approximate estimate of token consumption for documents of different sizes:

Actual token usage may vary depending on the structure and content of the file.