Meet the Models: Choose Your Tool

You've been running prompts across three models. This session is about making a deliberate choice, not just using whatever loads first.

What the three models are like

These are general tendencies as of mid-2026. They shift with model updates, and your own experience may differ.

Model Strength Watch for
Claude (claude.ai) Follows structured output instructions precisely; detailed CoT reasoning Can be verbose
Gemini (gemini.google.com) Strong JSON output; Canvas renders HTML live; fast
ChatGPT (chat.openai.com) Strong general performance Tends to wrap output in markdown fences even when told not to

These are tendencies, not rules. The comparison you're about to run will produce your own observations, which matter more.

Rate limits

All three platforms have free-tier rate limits. On a busy day you will hit them. The best model for a task that rate-limits you mid-build is not actually the best model in practice — the one that's available when you need it is.

That's why you're picking both a primary tool and a backup.

The comparison

Write a single complex prompt that uses at least three of the five techniques you've learned (zero-shot, one-shot, few-shot, structured output, chain-of-thought). Make it something you'd actually want to build for your Week 2 project.

Run it on all three models. For each run, record:

  • Did the model follow the structured output instructions?
  • Was the CoT reasoning (if used) useful?
  • How accurate was the output?
  • Did you hit a rate limit?
  • Which output needed the least editing?

Pick your tools

After running the comparison, write down:

Primary model for Week 2: ___ because ___.

Backup model: ___ (for rate limit situations).

Neither choice is permanent. Switch whenever it's useful. The point of this session is to go into the hackathon with a reasoned starting point rather than defaulting to whichever one you heard about first.