Models

Models is the catalogue of every AI model your company can use. Open it when you need to decide which model to work with:

Pick a model for a task. Each card describes what the model is good at and shows what it can do: how much input it accepts, whether it calls tools, reasons before answering or returns structured JSON.
Know what it costs before you use it. A model is either covered by your plan or billed per token. For a pay-as-you-go model, the card opens its providers with their prices, uptime and speed.
Call it from your own code. Pay-as-you-go models are the ones you call through the API; see Use the API.

Badges

The badges answer two questions before you use a model: what it costs, and what kind of model it is.

What it costs

Badge	What it means for you
Plan	Included in your subscription: usage counts against your plan's token allowance, so the card shows no price. These are Kartlos Premium and Kartlos Standard, the models of Kartlos, a multi-model AI system and orchestration model developed by DataMind. They work inside DataMind OS.
PAYG	Pay-as-you-go: you pay per token for what you use.

What kind of model

Badge	What it means for you
Premium	Kartlos Premium, suitable for complex tasks. It is the flagship model of DataMind OS: DataMind keeps it pointed at the strongest frontier model available; use it for complex reasoning, coding and agentic tasks. It draws on your Premium tokens.
Standard	Kartlos Standard, for everyday use. It is the fast everyday model of DataMind OS: lower latency for routine tasks, summaries and high-volume work. It draws on your Standard tokens.
Frontier, Pro, Lite	How capable DataMind rates the model: Frontier is the most capable class, Pro the middle one and Lite the lightest.
New	DataMind recently added the model to the catalogue.

Capability signals

The right side of each card shows what the model can do, so you can rule out a model before you try it:

Context window: how many tokens one request can hold, prompt and answer together. For long documents or long conversations, choose a larger one.
Tools: the model can call functions you define. You need this for agents and integrations.
Thinking: the model reasons before it answers. The tooltip shows whether thinking is mandatory or on by default, which effort levels DataMind allows for the model and the default one. To set the effort in an API request, see Use the API.
Structured output: the model can return JSON that follows a schema you provide, ready for your code to read.

Provider table

A pay-as-you-go model can be served by several providers. Open a pay-as-you-go card to compare them before you call the model. Prices are in USD per 1M tokens and already include any discount: a green price with a struck-through price above it is discounted.

Column	What it tells you
Provider	The provider that serves the model.
Input $/1M, Output $/1M	What you pay for the text you send and for the answer.
Cache read $/1M	What you pay for input the provider serves from its prompt cache, at a lower price than Input.
Uptime	How reliably the provider has been available recently.
Max output	The longest answer the provider returns in one response, in tokens.
Latency, Throughput	How quickly the provider has been responding recently and how fast it writes the answer, in tokens per second. Shown when data is available.

Provider sort in Settings decides which provider DataMind tries first.

Questions, answered

Where do I find a model's ID?

On the model's card, under the model name, for example anthropic/claude-opus-5.5. Send it in the model field of an API request.

Why do I see different models than a colleague in another company?

The list depends on your company's Region in Settings. A company set to EU sees only the models served in the EU.