Claude Fable 5
Anthropic's most capable model — the first of the Mythos-class tier, sitting above Opus. Built for the hardest reasoning and long-horizon writing: full-length feature structure, multi-episode continuity, dense research synthesis.
Max plan
Claude Fable 5 is available on the Max plan.
Capabilities
| Feature | Support |
|---|---|
| Chat (text-to-text) | Yes, streaming |
| Vision (image understanding) | Yes — up to 10 connected images |
| Document references | Yes — up to 100 text/document inputs |
| Context Window | 1,000,000 tokens |
| Max Output | 128,000 tokens |
| Reasoning Effort | Low / Medium / High / X-High — thinking is always on |
Always-On Thinking
Fable 5 reasons before every reply — there is no "no thinking" mode. The effort selector scales how deeply:
- Low — quick but still considered; fine for most writing passes.
- Medium / High — structural work: outlining a season arc, reconciling continuity across drafts.
- X-High — the heaviest setting available anywhere in NGMC; reserve it for problems where a wrong answer is expensive (final-draft structural review, complex world-logic audits).
When Fable Over Opus
| Task | Recommendation |
|---|---|
| Dialogue passes, rewrites, coverage | Opus 4.6 — faster, and excellent |
| Full-feature structure from a logline | Fable 5 |
| Multi-document synthesis (research + drafts + notes) | Fable 5 |
| Catching subtle logic/continuity breaks | Fable 5 at High/X-High |
Replies take longer than Opus — the always-on thinking is the price of the depth.
Limitations
- Max-plan exclusive
- Slower time-to-first-token than Opus 4.6 (thinking cannot be disabled)
- Text out only; no audio understanding
See Also
- Claude Opus 4.6 — the faster everyday writing partner
- GPT 5.5 — OpenAI's flagship alternative