# Billing for AI Studio

Krutrim Cloud's **AI Studio** offers access to a diverse catalogue of open-source and in-house AI models for text generation, embeddings, speech, and multimodal use cases. All models are billed **based on usage**, with costs varying depending on the type of input/output and the model in use.

***

### Model catalog billing

Each model in AI Studio defines:

* **Billing mode** (Input tokens, Output tokens, Audio minutes, etc.)
* **Unit of pricing** (Per 1M tokens, Per minute, Per hour, etc.)
* **Rate** (in ₹)

You are billed **only for what you use**. Charges are calculated based on usage metrics collected during API calls or console interactions with the models.

***

### Model Catalog Pricing

| Model Name                                    | Input Rate (₹)     | Output Rate (₹)    | Unit Type        | Task Type            |
| --------------------------------------------- | ------------------ | ------------------ | ---------------- | -------------------- |
| **DeepSeek-R1-Distill-Llama-70B**             | ₹10.00 / 1M tokens | ₹10.00 / 1M tokens | Tokens           | Text Generation      |
| **DeepSeek-R1-Distill-Llama-8B**              | ₹3.00 / 1M tokens  | ₹3.00 / 1M tokens  | Tokens           | Text Generation      |
| **DeepSeek-R1**                               | ₹11.00 / 1M tokens | ₹16.00 / 1M tokens | Tokens           | Text Generation      |
| **Google / Gemma-3-27b-it**                   | ₹8.00 / 1M tokens  | ₹25.00 / 1M tokens | Tokens           | Multimodal           |
| **Krutrim / Krutrim-1**                       | ₹16.6 / 1M tokens  | ₹16.6 / 1M tokens  | Tokens           | Text Generation      |
| **Krutrim / Krutrim-2**                       | ₹6.6 / 1M tokens   | ₹6.6 / 1M tokens   | Tokens           | Text Generation      |
| **Krutrim / Krutrim-TTS**                     | N/A                | ₹4.42 / min        | Minutes (output) | Text-to-Speech       |
| **Krutrim / Krutrim-Dhwani**                  | ₹24.00 / hour      | N/A                | Hours (input)    | Speech-to-Text       |
| **Krutrim / Bhasantarit**                     | ₹6.26 / 1M tokens  | N/A                | Tokens (input)   | Text-to-Embedding    |
| **Krutrim / Vyakyarth**                       | ₹6.06 / 1M tokens  | N/A                | Tokens (input)   | Text-to-Embedding    |
| **Krutrim / Chitrapathak**                    | ₹83.6 / 1M tokens  | ₹34.53 / 1M tokens | Tokens           | Image-to-Text        |
| **Krutrim / Tokenizer**                       | ₹0.00              | ₹0.00              | Tokens           | Tokenization Utility |
| **Meta / LLaMA-3.2-11B-Vision-Instruct**      | ₹14.94 / 1M tokens | ₹14.94 / 1M tokens | Tokens           | Image-to-Text        |
| **Meta / LLaMA-3.3-70B-Instruct**             | ₹73.04 / 1M tokens | ₹73.04 / 1M tokens | Tokens           | Text Generation      |
| **Meta / LLaMA-4-Maverick-17B-12BE-Instruct** | ₹17.00 / 1M tokens | ₹50.00 / 1M tokens | Tokens           | Multimodal           |
| **Microsoft / Phi-4 Reasoning Plus**          | ₹5.00 / 1M tokens  | ₹29.00 / 1M tokens | Tokens           | Text Generation      |
| **Mistral / Mistral-7B-v0.2**                 | ₹16.6 / 1M tokens  | ₹16.6 / 1M tokens  | Tokens           | Text Generation      |
| **Qwen / Qwen3-32B**                          | ₹8.00 / 1M tokens  | ₹25.00 / 1M tokens | Tokens           | Text Generation      |
| **Qwen / Qwen3-30B-A3B**                      | ₹8.00 / 1M tokens  | ₹25.00 / 1M tokens | Tokens           | Text Generation      |

***

### Additional Notes

* **Tokenization logic** varies by model and is handled automatically using the model’s default tokenizer.
* **Speech models** are priced per **minute/hour of audio**, based on the type (TTS or STT).
* **Multimodal models** may count both text and vision tokens.

***

### Evaluation Billing

The cost of running **Model Evaluations** and **Performance Evaluations** on the Krutrim platform is the **same as inference** — based on the number of tokens processed.

#### Token-Based Pricing

* You are billed per token used during evaluation.
* There are **no additional fees** for launching or running evaluation jobs.

#### Low Balance Handling

* If your account balance drops below a certain threshold, the platform will **automatically pause evaluation services** until funds are added.

### Fine-Tuning Billing

Krutrim offers transparent and flexible pricing for both **fine-tuning** and **deployment** of custom models. Below is a detailed breakdown of the billing model:

#### Fine-Tuning Pricing

Fine-tuning costs are calculated **per 1 million tokens**, and pricing varies based on the model used.

| Model               | Price (INR) | Unit          |
| ------------------- | ----------- | ------------- |
| Llama-3-8b-instruct | ₹33         | per 1M tokens |

> **Note:** This is **token-based pricing**, ensuring you pay only for the compute resources used during fine-tuning.

***

#### Deployment Pricing

Once your model is fine-tuned, you can deploy it on an **NVIDIA H100 GPU**. Pricing is based on actual GPU usage.

| GPU Resource | Price (INR) | Unit         |
| ------------ | ----------- | ------------ |
| NVIDIA H100  | ₹213        | per GPU-hour |

> Deployment pricing **includes inference usage** — you are charged only for the **active duration** of your deployment.

***

#### Cost Management Best Practices

* **Optimize Your Dataset**: High-quality, concise datasets reduce token usage and improve training efficiency.
* **Monitor Active Deployments**: Shut down unused deployments to avoid incurring extra GPU charges.
* **Use Checkpoints**: Save checkpoints during training to avoid repeating the entire process in case of interruptions.

***

#### Low Balance Handling

If your account balance falls below a defined threshold:

* **Fine-tuning and deployment services will be automatically paused.**
* Resume once sufficient balance is added to the account.

### Where to See This in Console

To see real-time usage and billing:

1. Go to **Billing → Usage → AI Studio**
2. Select the relevant sub-tab:
   * **Model Catalogue**
   * **Fine Tuning**
   * **Deployment**
   * **Evaluation**
3. Use the **date filter** to narrow your view
4. Click **Export** to download usage data


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.cloud.olakrutrim.com/basics/ai-studio/billing-for-ai-studio.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.