What Is Fable 5?
Anthropic's Most Powerful Public AI Model, Explained
Claude Fable 5 (API model ID: claude-fable-5) is the most capable model Anthropic has ever made generally available — the same training as the restricted Mythos 5, positioned above Opus, and state-of-the-art on nearly every public benchmark.
Fable 5 vs. Mythos: what's the relationship?
In April 2026, Anthropic previewed Mythos, its strongest internal model — but restricted it to a handful of partners because it was considered too skilled at finding and exploiting software vulnerabilities.
On June 9, Anthropic shipped that frontier as two products. Fable 5 is the public release: same training, frontier capabilities, plus safeguards designed for general availability. Mythos 5 is the same underlying model with safeguards lifted in some areas, available only to vetted cyberdefenders and infrastructure providers via Project Glasswing, in collaboration with the US government.
In one line: Fable 5 is the Mythos you can actually use.
👉 Deep dive: Claude Mythos 5 — what it is, who gets access, and the full Fable 5 comparison
How good is it?
| Benchmark | Fable 5 | Claude Opus 4.8 | GPT-5.5 |
|---|---|---|---|
| SWE-Bench Pro (coding) | 80.3% | 69.2% | 58.6% |
| AA Intelligence Index | 65 | Price-tier median: 36 | |
The longer and more complex the task, the larger the lead. Notable launch-day demonstrations:
- Stripe reported Fable 5 "compressed months of engineering into days"; in a 50-million-line Ruby codebase it completed a migration in one day that was estimated at two-plus months for a team.
- Beat Pokémon FireRed using vision alone — raw screenshots, no maps or helper harness. Earlier Claude models needed extensive scaffolding to play at all.
- Long-horizon memory: with file-based memory, its improvement at Slay the Spire was 3× larger than Opus 4.8's. Multi-day autonomous agent work is its biggest differentiator.
Caveat: most numbers are vendor-published or from early-access customers' own evals; independent third-party benchmarks are still limited this close to launch. Trade-offs: output speed ~60.3 tok/s and time-to-first-token ~81.7s are both worse than tier medians.
👉 Full head-to-head with GPT-5.5 and Opus 4.8, plus the one-paragraph verdict: Fable 5 review & comparison
Pricing: 2× Opus 4.8
$10 per million input tokens, $50 per million output — exactly double Opus 4.8's $5/$25. Cache hits cost $1/M; batch runs at half price ($5/$25); US-only inference adds a 1.1× multiplier. There is no flex tier, so for huge offline workloads GPT-5.5's $2.50/$15 flex pricing remains cheaper.
Rule of thumb: use the cheapest model that reliably clears your quality bar. Fable 5 earns its premium on long, complex, compounding work — and on hard tasks it can use fewer tokens and less rework, narrowing or even inverting the effective cost gap.
👉 Full price table, free-window timeline, and an interactive calculator: Fable 5 pricing
The safety-classifier fallback (know this before you build)
Fable 5 ships with conservatively tuned classifiers. Queries touching cybersecurity, biology, chemistry, or large-scale distillation are answered by Opus 4.8 instead — you're notified, and that request is billed at Opus 4.8 rates. Anthropic says this triggers in under 5% of sessions and will sometimes catch harmless requests.
Compliance note: using Fable 5 requires 30-day data retention for safety monitoring (not used for training; human access is logged).
How to get access
① claude.ai subscriptions (limited-time window)
From June 9–22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost, counting as 2× usage. On June 23 it's removed from those plans; continued use requires usage credits. Claude.ai and Claude Code how-tos: fableclaude.com.
② API & cloud platforms
Generally available on the Claude API, AWS Bedrock, Vertex AI, and Microsoft Foundry as claude-fable-5. Developer quickstart: fableapi.app.
Skip the setup — start using Fable 5 now
OmniaKey: Fable 5 at $3/$15 (limited-time, ~70% off) · official-compatible endpoint · no card required
FAQ
Is Fable 5 free?
Through June 22, 2026 it's included in paid claude.ai plans (at 2× usage). From June 23 it requires usage credits. The API is always metered at $10/$50 per million tokens.
Is Fable 5 better than GPT-5.5?
It leads decisively on coding (SWE-Bench Pro 80.3% vs 58.6%) and overall intelligence scores, but it's slower, has very high first-token latency, and costs more per token.
Fable 5 or Opus 4.8?
Long-horizon agents, huge codebases, and 1M-token contexts justify Fable 5's 2× price. For routine tasks, Opus 4.8 (or Sonnet 4.6) is the better economics.