How LLMs Differ

Not All Language Models Are the Same

In the previous lessons, you learned that the LLM is the brain of every AI agent. But just like people have different strengths, different language models excel at different things. Choosing the right LLM for your task can make a significant difference in the quality of your results.

This lesson gives you a practical understanding of the major LLMs you will encounter in this course and when each one shines.

Claude (by Anthropic)

Claude is known for producing high-quality, thoughtful responses that follow instructions precisely. It is the model behind Claude Code, one of the primary agent platforms covered in this course.

Where Claude excels:

Writing quality. Claude produces clean, well-structured prose. It is particularly good at matching a requested tone, whether that is professional, casual, technical, or creative.
Following detailed instructions. If you give Claude a long, specific set of requirements, it tends to follow them faithfully without skipping steps or taking shortcuts.
Analysis and reasoning. Claude performs well on tasks that require understanding complex information, comparing options, and drawing conclusions.
Safety and honesty. Claude is designed to be transparent about its limitations. It will tell you when it is unsure rather than making something up.

Agent platform: Claude Code gives the agent direct access to your file system and terminal, making it powerful for tasks that involve creating, editing, and organizing files on your computer.

ChatGPT (by OpenAI)

ChatGPT is the most widely known AI model and has the largest ecosystem of integrations and plugins. It is versatile and works well across a broad range of tasks.

Where ChatGPT excels:

Versatility. ChatGPT handles a wide range of tasks competently, from creative writing to data analysis to coding help. It is a strong generalist.
Plugin ecosystem. ChatGPT has access to thousands of third-party plugins that extend its abilities. You can connect it to Zapier, Canva, Wolfram Alpha, and many other services.
Image generation. ChatGPT can create images through DALL-E integration, which is useful for visual content like social media graphics, logos, and illustrations.
Widespread adoption. Because ChatGPT is so popular, there are extensive community resources, tutorials, and shared prompt libraries available.

Agent platform: ChatGPT’s “GPTs” feature (requires a paid plan) lets you create custom agents with specific instructions, knowledge files, and tool access. All users can use Custom Instructions for persistent settings. The web browsing and code interpreter tools turn it into a capable agent.

Gemini (by Google)

Gemini is Google’s AI model, and its standout feature is deep integration with the Google ecosystem and strong multimodal capabilities.

Where Gemini excels:

Multimodal understanding. Gemini can process and reason about images, video, and audio, not just text. You can show it a photo and ask questions about it, or have it analyze a video.
Google integration. Gemini works natively with Gmail, Google Docs, Google Sheets, Google Calendar, and other Google Workspace tools. If your workflow is built on Google services, this integration is a major advantage.
Long context. Gemini can handle very long documents and conversations. If you need to analyze a 100-page report or maintain context across a lengthy interaction, Gemini handles this well.
Real-time information. Gemini has access to Google Search, giving it up-to-date information about current events, recent publications, and live data.

Agent platform: Google’s AI Studio and Gemini API let you build agents that leverage Google’s services. Gemini within Google Workspace can act as an agent for email, document, and spreadsheet tasks.

Ollama and LM Studio (Local Models)

Ollama and LM Studio are not specific models but rather platforms that let you run open-source language models on your own computer. This is a fundamentally different approach from the cloud-based options above.

Where local models excel:

Privacy. Your data never leaves your computer. Nothing is sent to a company’s servers. For sensitive personal or business information, this can be critical.
No usage costs. Once you download a model, you can use it as much as you want without paying per message or per token. There are no monthly subscription fees.
Offline access. Local models work without an internet connection. You can use them on a plane, in areas with poor connectivity, or in environments where internet access is restricted.
Customization. You can choose from hundreds of open-source models, each with different strengths, sizes, and specializations.

Important trade-off: Local models require a reasonably powerful computer. A laptop with at least 16 GB of RAM can run smaller models, but the largest and most capable models need more powerful hardware. The quality of local models has improved dramatically, but the best cloud models (Claude, ChatGPT, Gemini) still outperform most local options on complex reasoning tasks.

When to Use Which

Here is a practical guide to choosing the right LLM for your task:

Scenario	Best Choice	Why
Writing blog posts, articles, or reports	Claude or ChatGPT	Both produce high-quality written content with good structure
Tasks involving Google Docs, Gmail, Sheets	Gemini	Native integration means seamless workflow
Creating images alongside text	ChatGPT	Built-in DALL-E image generation
Analyzing photos, videos, or screenshots	Gemini	Strongest multimodal capabilities
Working with sensitive or private data	Ollama / LM Studio	Data stays on your computer
Following very specific, detailed instructions	Claude	Excels at precise instruction-following
Experimenting on a budget	Ollama / LM Studio	No per-message costs after setup
General-purpose tasks	Any of them	All major models handle common tasks well

The Key Insight

Here is the most important takeaway from this lesson: the agent pattern works the same regardless of which LLM powers it. The core loop of Observe, Think, Act applies to all of them. The four components (LLM, Tools, Loop, Memory) are present in all agent platforms. What changes is the specific strengths of the brain and the specific tools available.

This means that the skills you learn in this course are transferable. If you learn to build agents with Claude and later want to try ChatGPT or Gemini, you will not be starting over. The principles are the same. Only the interface and some specific features are different.

Free Tiers vs. Paid Tiers

Most platforms offer a free tier that lets you get started without spending any money:

Claude: The free tier provides access to Claude with limited usage per day. The Pro plan offers higher limits and access to the most capable models.
ChatGPT: A free tier provides access to ChatGPT’s current standard model with limited usage. ChatGPT Plus ($20/month) gives you higher limits, faster responses, and access to advanced features.
Gemini: A free tier is available with Google account access. Google One AI Premium provides higher usage limits and deeper Workspace integration.
Ollama / LM Studio: Completely free to use. The only cost is the hardware you run them on, which you likely already own.

What Comes Next

Now that you understand what agents are, how the core loop works, what the four components are, and how different LLMs compare, you have a solid foundation. The next section of the course dives into prompts and configuration, where you will learn how to give agents clear instructions that produce excellent results. That is where you start building real things.