Do AI models have personalities?
Dr Susanne Friese · · Experiments & evidence
Do AI Models Have Personalities? Why Choosing the Right One Matters
You may have heard about the latest addition to the Claude 3 family—Claude 3.7—released on February 24, 2025.
Claude 3.7 is one of the new reasoning models, featuring an extended thinking mode (available to pro users). When activated, this mode allows Claude to take more time before responding, supposedly improving the quality of answers that require deeper reasoning. At least, that's the idea.
What Are Reasoning Models? Unlike standard AI models that generate quick responses based on patterns, reasoning models are designed to “think” step by step before answering. This means they analyze the problem more thoroughly before providing a response. 🔹 They can break down complex problems into smaller steps—a method known as Chain of Thought reasoning. 🔹 They can refine their answers internally, reducing the risk of incorrect or misleading responses. 🔹 They trade speed for accuracy, meaning responses may take longer but should be more reliable—though this depends on execution.
From my experience with OpenAI’s reasoning models (o1 and o3-mini), I’ve seen a significant improvement in response quality, and I’m impressed by what’s possible now compared to just last summer.
However, initial reactions to Claude 3.7 seem more mixed. Here are some first impressions from power users, as reported by The Neuron
meco.app/share/67354690-5598-43ee-bd61-053… (March 3, 2025): “3.7 is an overdesigning machine. One poor soul watched their tidy 400-line script explode into a 1,100-line monster. More isn't always better, Claude!” “It has selective hearing—ignoring instructions more often than a teenager being asked to clean their room.” “The personality downgrade is real. 3.7 apparently has the conversational charm of a wooden table.” “It suffers from complexity amnesia. While it starts strong, it gets lost in the sauce on longer, more complex projects.” One user summed it up this way: “3.7 feels like a senior McKinsey consultant—presenting confidently while completely missing the point.” 😱
Despite the criticisms, how you interact with the model matters. Here’s what experienced users recommend:
✅ Provide a detailed initial prompt—Avoid short, back-and-forth interactions and instead lay out clear instructions upfront. ✅ Be highly specific—Micromanage constraints and requirements to prevent the AI from wandering off course. ✅ Set clear expectations—Clearly define what you want the AI to do, how detailed the response should be, and any formatting guidelines.
The Takeaway Each AI model has its own personality—interacting with different models can feel like talking to entirely different people. Of course, these are machines, but behind them are teams of developers who have programmed their behavior through system prompts and specific instructions.
This is why it truly matters which model you use. The way they process and respond to your requests isn’t just about raw intelligence—it’s shaped by the way they’ve been designed to "think." Understanding these differences can help you choose the right model for the right task.
If you haven’t experimented with different models yet, why not set it as a goal for this week or month? Here are a few links to help you find the best model for your task:
Claude
Gemini - chat to supercharge your ideas
Perplexity (good for research)
The good old chatGPT
Le Chat (the French way) - Mistral AI
Happy experimenting. And as always, have a good start into your week,