What this course is
Claude Fable 5 is the strongest—and most expensive—model in Anthropic’s current lineup. It became available globally again on July 1, 2026, after a temporary suspension related to U.S. government export controls. This course teaches you to prompt this specific model well, not using Claude in general.
The content cross-checks a video lesson with 6 practical techniques against Anthropic’s official documentation, citing the official text whenever there’s a verifiable citation — and clearly marking when something is general prompt engineering practice, not a literal quote from the Fable 5 docs.
💰 Official pricing (platform.claude.com/docs/pricing)
| Model | Input | Cache hit | Output |
|---|---|---|---|
| Claude Fable 5 / Claude Mythos 5 | $10/MTok | $1/MTok | $50/MTok |
| Claude Opus 4.8 | $5/MTok | — | $25/MTok |
| Claude Sonnet 5 (through Aug 31, 2026) | $1/MTok | — | $5/MTok |
| Claude Haiku 4.5 | $0.50/MTok | — | $2.50/MTok |
Fable 5 costs exactly twice as much as Opus 4.8 (input and output)—the most expensive in the catalog. Batch API (50% off) brings Fable 5 down to $5/$25 per MTok. For context: a 1M-token window by default, with no “premium” for long context.
🗓️ Quick context: why this course now
Fable 5 and Mythos 5 launched on June 9, 2026. On June 12, the U.S. government imposed export controls on both models, and because nationality couldn’t be verified in real time, Anthropic suspended access for ALL users. The real reason: Amazon researchers reported a way to bypass Fable 5’s safeguards—but Anthropic’s own test confirmed that much less capable models (Opus 4.8, GPT-5.5, Kimi K2.7, and even Haiku 4.5, Sonnet 4.6, and Opus 4.6/4.7) could produce the same exploit demonstration. This wasn’t a capability unique to Fable 5—the suspension was a regulatory precaution. On July 1, 2026 (today), both models are returning globally.
Capabilities officially highlighted (vs. Opus 4.8)
Source: official “Introducing Claude Fable 5” doc, comparing it with Opus 4.8. Early adopter reports (unofficial): Dan Shipper/Every scored it 91/100 on the “Senior Engineer” benchmark (compared with Opus 4.8’s 63); Boris Cherny (Anthropic) publicly called it “the best model I’ve ever used for code, by a wide margin.”
💡 How I decide which model to use
This corresponds to how I manage the process day to day. I use the Opus for the architecture stage, the Sonnet when the task is well defined, and Fable when the task is complex and needs its own investigation cycle. The calling feature for OpusPlan is underrated: most people never use it explicitly, then wonder why Opus doesn’t switch over automatically.
How the course works
2 tracks, 14 modules, 84 topics. Track 1: the 6 techniques from a video lesson, each cross-referenced with Anthropic’s official documentation, plus a bonus module with extra tricks found only in the docs. Track 2: 12 practical tips for using Fable 5 as a work agent — the right task, depth, limits, verification, subagents, memory, and loops — plus a bonus module with the master prompt.
✓ What this course covers
- ✓The 6 prompting techniques from the lesson, each paired with an official citation when available
- ✓Effort levels (max / xhigh / high / medium / low) and when to use each
- ✓The 4 security refusal categories and ways to configure fallback to Opus 4.8
- ✓Official extras: subagents, memory system, and the send_to_user tool
✗ What this course is NOT
- ✗This isn’t about the geopolitical export control controversy—that’s context for one paragraph, not the central topic
- ✗This isn’t a generic “how to use Claude” course—it’s specifically about prompting for Fable 5
- ✗Don’t invent numbers or capabilities beyond what Anthropic has documented
📖 Prompting Claude Fable 5
Context, official prompting techniques, effort levels, safety routing, and documentation extras—all to get the most out of Anthropic’s strongest model.
🧠 Fable 5 as a Work Agent
From one-off answers to a work system: 12 practical tips for using Fable 5 as a senior agent—choosing the right task, depth, limits, verification, subagents, memory, and loops—plus a bonus module with the master prompt.