Hands on Grok 2 Model: 'Grok 2 is Going to be Open Source'

Grok is an advanced AI model developed by xAI, the artificial intelligence company founded by Elon Musk. Unlike many mainstream language models, Grok is designed to be less restricted and more open in its responses compared to ChatGPT (by OpenAI) or Claude (by Anthropic). It aims to provide an unbiased, truth-seeking AI experience, making it one of the most powerful and distinctive large language models (LLMs) available today. With the Grok model such as Grok 2, you can explore vision capabilities, text to text and more. Grok 2, like its predecessors, is trained to engage in dynamic conversations, answer complex questions, and even generate creative content while maintaining a logical and fact-based approach.

Table of contents

• The Family of Grok by Elon Musk
• What is Grok-2?
• The Truth About Grok’s Image Generation
• The Release of Grok 3: Why This Changes Everything
• How to Access Grok Models via API?
• Hands-on Grok 2 Model
• Image Understanding: Vision model
• Grok 2 Using LangChain
• Let’s create a Chatbot
• Conclusion

The Family of Grok by Elon Musk

Here’s the evolution:

• Grok-1: Launched in November 2023, this initial version laid the groundwork for Grok’s conversational abilities. It was described as a “very early beta product,” with limited training but potential for rapid improvement.
• Grok-1.5: Released in May 2024, this update enhanced reasoning capabilities and increased the context length to 128,000 tokens, allowing for more complex interactions.
• Grok-1.5V: Announced in April 2024, this version added the ability to process various visual information types, such as documents and images.
• Grok-2: Debuted in August 2024, Grok-2 featured upgraded performance and introduced image generation capabilities, further expanding its functionality.
• Grok-2 mini: Also released in August 2024, this smaller version aimed to balance speed and answer quality, catering to users with different needs.
• Grok-3: The latest iteration, unveiled on February 18, 2025, emphasizes complex reasoning and high-level problem-solving capabilities. This version is seen as a significant leap forward in AI technology and aims to surpass existing chatbots in performance.

Now, Grok 3 is here!!!

Read the full coverage here: Grok 3 is Here! And What It Can Do Will Blow Your Mind!

What is Grok-2?

Grok-2 is an enhanced version of the Grok chatbot developed by Elon Musk’s xAI. It is designed to be more user-friendly, adaptable, and proficient in answering questions, assisting with writing, and tackling coding challenges.

Key Features and Capabilities:

• Accessibility Grok-2 is available to X (formerly Twitter) Premium subscribers.
• Performance xAI claims Grok-2 is a “notable advancement” over its predecessor, Grok-1.5, with upgraded performance and reasoning. An updated version of Grok-2 is reportedly three times faster with improved accuracy, instruction-following, and multilingual capabilities.
• Ranking Grok-2 holds the third position on the LMSYS leaderboard under the identifier ‘sus-column-r,’ outperforming Claude 3.5 from Anthropic, Gemini, and Llama 3.
• Image Generation Grok-2 has image-creation abilities.
• Real-time Data Grok-2 uses real-time data sourced from X posts.
• Grok Button The “Grok button” helps users discover relevant context, understand real-time events, and delve deeper into trending discussions.
• Versions xAI has launched a smaller, quicker version called Grok-2 mini that balances the speed and quality of responses.

As per Elon Musk, Grok-2 will be open-source in the coming days.

The Truth About Grok’s Image Generation

One common misconception is that Grok 2 generates images—but it doesn’t. Instead, xAI relies on a separate image model called Flux.

• Flux is one of the most advanced AI image-generation models, especially when it comes to realistic outputs, surpassing even MidJourney.
• When users request an image on X, Grok acts as an intermediary, calling Flux via an API to generate the image.

So, while it may seem like Grok is directly creating images, it’s actually leveraging Flux behind the scenes.

The Release of Grok 3: Why This Changes Everything

Now is the perfect time to start building AI agents with the Grok API because Grok 3 is here—and it might be the best AI model in the world.

Note: In one or two you can access the Grok 3 through API.

Why is Grok 3 so important?

• Unprecedented Compute Power: xAI now has the largest AI cluster in the world, and it’s still expanding.
• First-Principles Thinking: Musk and his team approach AI from a fundamentals-first perspective, ensuring maximum efficiency and performance.
• Advanced Reasoning: Grok 3 excels in solving complex problems with innovative logic, outperforming existing models.
• DeepSearch: Enables rapid, in-depth searches with the option to focus on specific sources.
• Big Brain: Enhances response depth by allocating more time for thoughtful analysis.
• Responsible AI: First chain-of-thought model from x.AI, ensuring transparency and safeguards against bias and misinformation.
• Speed & Power: Runs on the Colossus Supercomputer, three times faster than Grok 2, ideal for real-time applications.
• Voice & Audio: Upcoming features include voice mode and audio-to-text conversion.

Most people are still using ChatGPT or Claude, meaning Grok is still under the radar. But if Grok 3 proves superior in the coming days, it could rapidly skyrocket in popularity, potentially becoming the world’s leading LLM within a year or two.

How to Access Grok Models via API?

As of now, you can access two models:

• grok-2-1212: The flagship LLM that delivers unfiltered insights and raw intelligence.
• grok-2-vision-1212: The latest image-understanding LLM that excels at processing diverse visual inputs like documents and photos.

Firstly, open: https://x.ai/

Click on “Start Building Now” and sign in using your email id.

Add payment method and billing address as required.

This will set your API key for the model you want to use.

Hands-on Grok 2 Model

Save the API key in the .env file like: GROK_API_KEY = “Your_API_KEY”

import os
from openai import OpenAI
From dotenv import load_dotenv
load_dotenv()
GROK_API_KEY=os.getenv(“GROK_API_KEY”)
client = OpenAI(
api_key= GROK_API_KEY,
base_url="https://api.x.ai/v1",
)
completion = client.chat.completions.create(
model="grok-2-latest",
messages=[
{"role": "system", "content": "You are a Python progamminmg language expert."},
{"role": "user", "content": "How can a beginner read and parse a JSON file in Python? You can teach them using examples"},
],
)
print(completion.choices[0].message.content)

Output

To help a beginner read and parse a JSON file in Python, let's go through the
 process step-by-step with examples. We'll use the `json` module, which is
 part of Python's standard library, making it easy to work with JSON data.

Here's how you can read and parse a JSON file:

1. First, we need to import the `json` module.

2. Then, we'll open the JSON file and use `json.load()` to parse its contents.

3. Finally, we can access and use the parsed data.

Let's see this in action with an example:

```python

# Import the json module

import json

# Open and read the JSON file

with open('example.json', 'r') as file:

# Parse the JSON data

data = json.load(file)

# Now we can work with the data

print(data)

# Accessing specific elements (assuming the JSON structure)

print(data['name'])

print(data['age'])

# If it's a list of objects

for item in data:

print(f"Name: {item['name']}, Age: {item['age']}")

```

Let's break down what's happening:

1. We import the `json` module, which provides functions to work with JSON
 data.

2. We use a `with` statement to open the file. This ensures the file is
 properly closed after we're done with it.

3. We use `json.load(file)` to read and parse the JSON data from the file.
 This function automatically converts the JSON data into a Python object

 (usually a dictionary or a list).

4. After parsing, `data` now contains the JSON content as a Python object. We
 can print it or access its elements.

5. Depending on the structure of your JSON, you can access elements using
 dictionary syntax (`data['key']`) or list indexing.

Here's an example of what the `example.json` file might look like:

```json

{

"name": "John Doe",

"age": 30,

"city": "New York"

}

```

Or, if it's a list of objects:

```json

[

{"name": "John Doe", "age": 30},

{"name": "Jane Smith", "age": 25}

]

```

For beginners, here are some important points to remember:

- JSON files typically have a `.json` extension.

- JSON data can be a single object (like a dictionary) or an array (like a
 list).

- You need to know the structure of your JSON to access its elements
 correctly.

- If you encounter errors, it's often because the JSON file is not properly
 formatted or you're trying to access elements that don't exist.

To practice, beginners can create their own JSON files and try parsing them.
 They can also use online JSON validators to ensure their files are correctly
 formatted before attempting to parse them in Python.

if completion.usage:
print(completion.usage.to_json())

{<br>"completion_tokens": 713,<br>"prompt_tokens": 40,<br>"total_tokens": 753,<br>"prompt_tokens_details": {<br>"audio_tokens": 0,<br>"cached_tokens": 0,<br>"text_tokens": 40,<br>"image_tokens": 0<br>}<br>}

Cost of Grok 2 Output

Cost of Input (Prompt Tokens)

• 40 tokens × ($2.00 / 1,000,000)
  = $0.00008

Cost of Output (Completion Tokens)

• 713 tokens × ($10.00 / 1,000,000)
  = $0.00713

Total Cost for This Request

So, this request costs approximately $0.00721 (or about 0.72 cents).

Image Understanding: Vision model

import os
from openai import OpenAI
From dotenv import load_dotenv
load_dotenv()
GROK_API_KEY=os.getenv(“GROK_API_KEY”)
client = OpenAI(
api_key= GROK_API_KEY,
base_url="https://api.x.ai/v1",
)
completion = client.chat.completions.create(
model="grok-2-latest",
messages=[
{"role": "system", "content": "You are a Python progamminmg language expert."},
{"role": "user", "content": "How can a beginner read and parse a JSON file in Python? You can teach them using examples"},
],
)
print(completion.choices[0].message.content)

To help a beginner read and parse a JSON file in Python, let's go through the
 process step-by-step with examples. We'll use the `json` module, which is
 part of Python's standard library, making it easy to work with JSON data.

Here's how you can read and parse a JSON file:

1. First, we need to import the `json` module.

2. Then, we'll open the JSON file and use `json.load()` to parse its contents.

3. Finally, we can access and use the parsed data.

Let's see this in action with an example:

```python

# Import the json module

import json

# Open and read the JSON file

with open('example.json', 'r') as file:

# Parse the JSON data

data = json.load(file)

# Now we can work with the data

print(data)

# Accessing specific elements (assuming the JSON structure)

print(data['name'])

print(data['age'])

# If it's a list of objects

for item in data:

print(f"Name: {item['name']}, Age: {item['age']}")

```

Let's break down what's happening:

1. We import the `json` module, which provides functions to work with JSON
 data.

2. We use a `with` statement to open the file. This ensures the file is
 properly closed after we're done with it.

3. We use `json.load(file)` to read and parse the JSON data from the file.
 This function automatically converts the JSON data into a Python object

 (usually a dictionary or a list).

4. After parsing, `data` now contains the JSON content as a Python object. We
 can print it or access its elements.

5. Depending on the structure of your JSON, you can access elements using
 dictionary syntax (`data['key']`) or list indexing.

Here's an example of what the `example.json` file might look like:

```json

{

"name": "John Doe",

"age": 30,

"city": "New York"

}

```

Or, if it's a list of objects:

```json

[

{"name": "John Doe", "age": 30},

{"name": "Jane Smith", "age": 25}

]

```

For beginners, here are some important points to remember:

- JSON files typically have a `.json` extension.

- JSON data can be a single object (like a dictionary) or an array (like a
 list).

- You need to know the structure of your JSON to access its elements
 correctly.

- If you encounter errors, it's often because the JSON file is not properly
 formatted or you're trying to access elements that don't exist.

To practice, beginners can create their own JSON files and try parsing them.
 They can also use online JSON validators to ensure their files are correctly
 formatted before attempting to parse them in Python.

Input image

Output

if completion.usage:
print(completion.usage.to_json())

{<br>"completion_tokens": 713,<br>"prompt_tokens": 40,<br>"total_tokens": 753,<br>"prompt_tokens_details": {<br>"audio_tokens": 0,<br>"cached_tokens": 0,<br>"text_tokens": 40,<br>"image_tokens": 0<br>}<br>}

import os
from openai import OpenAI
from getpass import getpass
XAI_KEY = getpass('Enter Grok API Key: ')

Cost of Grok 2 Output

Cost of Input (Prompt Tokens)

• Text input cost:
  • 52 tokens × ($2.00 / 1,000,000) = $0.000104
• Image input cost:
  • 768 tokens × ($2.00 / 1,000,000) = $0.001536
• Total input cost = $0.00164

Cost of Output (Completion Tokens)

• Text output cost:
  • 483 tokens × ($10.00 / 1,000,000) = $0.00483

Total Cost for this request

So, this request costs approximately $0.00647 (or about 0.65 cents).

Grok 2 Using LangChain

os.environ['XAI_API_KEY'] = XAI_KEY
image_url = (
"https://safarisafricana.com/wp-content/uploads/2020/05/Copy-of-SAFAF-2X1-images-60.jpg"
)
client = OpenAI(
api_key=XAI_KEY,
base_url="https://api.x.ai/v1",
)
messages = [
{
"role": "user",
"content": [
{
"type": "image_url",
"image_url": {
"url": image_url,
"detail": "high",
},
},
{
"type": "text",
"text": """Please provide a detailed description of the contents in this image, including any notable objects, colors, patterns, and the overall scene.
If there are any texts or symbols, please transcribe or interpret them as well.""",
},
],
},
]
completion = client.chat.completions.create(
model="grok-2-vision-latest",
messages=messages,
temperature=0.01,
)
print(completion.choices[0].message.content)

Output

The image depicts a vibrant and lush savanna scene, showcasing a variety of<br> wildlife in their natural habitat. Here is a detailed description:<br><br>### Animals:<br><br>1. **Giraffes**: There are three giraffes in the image, standing tall with<br> their long necks and distinctive spotted patterns. Their colors are<br> primarily light brown with orange-brown patches. They are positioned towards<br> the center and right side of the image.<br><br>2. **Zebras**: Several zebras are present, identifiable by their black and <br>white striped patterns. They are scattered across the scene, with some<br> standing and others grazing. Their stripes create a striking contrast <br>against the green grass.<br><br>3. **Antelopes/Deer**: There are multiple antelopes or deer-like animals,<br> with slender bodies and light brown coats. Some have white underbellies and<br> legs. They are smaller in size compared to the giraffes and zebras, and are<br> seen grazing or standing around.<br><br>### Vegetation:<br><br>- **Grass**: The ground is covered with lush green grass, indicating a<br> healthy and vibrant ecosystem.<br><br>- **Bushes and Trees**: There are various green bushes and small trees<br> scattered throughout the scene. In the background, there are larger trees,<br> including one with a broad canopy on the left side of the image.<br><br>### Colors:<br><br>- **Green**: Dominant color due to the grass, bushes, and trees.<br><br>- **Brown**: Seen in the giraffes' patches, antelopes' coats, and some parts<br> of the zebras.<br><br>- **Black and White**: The zebras' stripes.<br><br>- **Orange**: Subtle tones in the giraffes' patches.<br><br>### Patterns:<br><br>- **Giraffe Spots**: Irregular, orange-brown patches on a light brown<br> background.<br><br>- **Zebra Stripes**: Bold, black and white stripes with varying thickness.<br><br>- **Antelope/Deer**: Light brown with white underbellies and legs.<br><br>### Overall Scene:<br><br>The scene is set in a savanna or grassland environment, typical of African<br> landscapes. The animals are coexisting peacefully, suggesting a harmonious<br> ecosystem. The lighting suggests it might be daytime, with natural sunlight<br> illuminating the scene, enhancing the vividness of the colors.<br><br>### Texts or Symbols:<br><br>There are no visible texts or symbols in the image.<br><br>This image captures the essence of wildlife in a natural setting,<br> highlighting the diversity and beauty of the animals and their environment.

Let’s create a Chatbot

if completion.usage:
print(completion.usage.to_json())

Output

{<br>"completion_tokens": 483,<br>"prompt_tokens": 820,<br>"total_tokens": 1303,<br>"prompt_tokens_details": {<br>"audio_tokens": 0,<br>"cached_tokens": 0,<br>"text_tokens": 52,<br>"image_tokens": 768<br>}<br>}

Conclusion

With Grok 3’s launch approaching, xAI is rapidly positioning itself as a serious challenger to OpenAI, Google DeepMind, and Anthropic. If xAI’s massive computing power translates into a superior AI model, Grok could dominate the AI space in the next few years. For developers, entrepreneurs, and AI enthusiasts—now is the time to start building with Grok. Early adopters stand to gain a major advantage in what could be the biggest AI revolution yet.

Experience the power of xAI Grok 3, the smartest AI on Earth! Enroll in our course to explore its groundbreaking features and transform your projects today!

I hope you found this article on Grok 2 hands-on informative. Let me know your thoughts in the comment section below.

The above is the detailed content of Hands on Grok 2 Model: 'Grok 2 is Going to be Open Source'. For more information, please follow other related articles on the PHP Chinese website!