1.1 Module 1 · The New Claude

Four Tiers, One Heuristic

Forget "which is better, Claude or ChatGPT." In 2026 the real question is which Claude, doing which job. Here's how it works, the four-model ladder, the one rule for choosing — and a 72-hour cautionary tale.

Which Claude? Picker Model Comparison

How Claude actually works (90 seconds)

Claude is a prediction engine trained on a vast amount of text — not a database, not a search engine. It generates the most plausible next words given everything in front of it. That one idea explains almost everything else.

Brilliant at

Language work — drafting, summarising, reformatting, explaining, reasoning over the material you give it.

Where it guesses

Recalling things it was never given. That's hallucination — confident gap-filling, not lying. The fix: give it the source and verify what matters.

In 2026 Claude also uses tools — web search, code, your files, your apps — which removes much of the old weakness. But the underlying behaviour is the same: give it good input and a clear job, check the bits that matter.

Interactive: Which Claude? Picker

The four-tier ladder — each step trades cost and speed for capability. Pick a task below and watch the right model light up.

What's the job?

The one rule to remember

Start on Sonnet. Move up to Opus when the task is hard or the stakes are high. Drop to Haiku when you just need speed and volume.

Model choice is gears. You don't drive everywhere in fifth — Sonnet for the commute, Opus for the hills, Haiku when you just need to move fast and cheap.

The cautionary tale: Fable 5

9 June 2026

Anthropic launches Fable 5, a Mythos-class model — state-of-the-art on most benchmarks, the most capable model in the world.

12 June 2026

Suspended for all customers after a US government export-control directive. Opus 4.8 and everything else stayed up.

The business lesson: frontier-model access can be revoked overnight by regulators. That's not a reason to avoid frontier AI — it's a reason to never depend on a single model. Build with fallback models in mind.

Key takeaway

Sonnet by default, Opus when it's hard, Haiku for volume — and never build on one model alone.