I’ve installed several artificial intelligence (“AI”) models on my own computers; I’m exploring the quality of responses from local AI and comparing the results.  This test is simply this a screenshot and a prompt.  This is the screenshot:

This is the prompt:

What can you tell me about this screenshot? what is in it and what does it accomplish?

The screenshot is from a program called Fritzing which let’s you model a breadboard circuit in a diagram.  The circuit depicted here is a power rail that takes the small positive and negative values from the geophone and bumps them up to the median voltage of the Analog to Digital circuit’s (“ADC”) voltage boundaries: 0 – 3.2V.  Raising the geophone’s voltages by 1.6V allows the ADC to see the negative voltages; otherwise values below 0 would not be within the ADC’s specifications and the results would not be reliable.

I tried with two local models which I refer to as Qwen and DeepSeek.

Here is the Qwen response with some follow-up questions & answers.

Here is the DeepSeek response.  I was disappointed with the result and decided not to ask further questions.

Here is the ChatGPT response.  This provided a higher level analysis and correctly concludes it is an electronic seismometer.

Conclusion

I prefer the Qwen response, it actually analyzed the circuit and rendered an opinion, whereas the DeepSeek simply described what it saw.