Heads up: some links below are referral links. They cost you nothing extra and never change what we recommend. How this site makes money.
The short answer
Claude Fable 5.1 is Anthropic's new top model, released September 1, 2026. It leads
the published benchmarks for coding, knowledge work and agentic research, and it's available now on
Claude, Claude Code and the API (claude-fable-5-1).
The headline isn't just capability. Per-token prices are unchanged from Fable 5 ($10 per million input, $50 per million output), but cache reads dropped 75% — Anthropic estimates typical workloads cost ~25% less, and heavily agentic ones up to ~45% less.
Should you switch? On a subscription, Fable is still the heaviest drain on your usage allowance — Sonnet 5 remains the sensible default for everyday work. Fable 5.1 earns its cost on long agentic runs, hard debugging, and research-grade problems.
| Benchmark | Fable 5.1 | Fable 5 | Opus 5 |
|---|---|---|---|
| Terminal-Bench-Science 0.1 (agentic research) | 52.6% | 24.7% | 29.0% |
| Terminal-Bench 4.0 (agentic coding) | 55.8% | 42.0% | 52.3% |
| Humanity's Last Exam (no tools) | 60.9% | 57.8% | 56.6% |
| CursorBench 3.2.0 | 73.4% | 70.5% | 70.0% |
| GDPval-AA v2 (knowledge work) | 1853 | 1723 | 1824 |
What's actually new
Three things stand out from Anthropic's announcement, beyond the benchmark table above.
Effort levels matter now. Fable 5.1 defaults to High effort in Claude Code and Medium in the Claude app. At Low or Medium effort it matches or beats Fable 5's results at a much lower cost — so the same model spans "cheaper than the old one" to "new state of the art" depending on how you set it.
The cache-read price cut is the quiet story. Cache reads — the model re-reading context it has already processed — now cost $0.25 per million tokens, down 75%. Agentic work is dominated by exactly this kind of re-reading, which is where the up-to-45% savings figure comes from. If you build on the API, this changes the economics more than any benchmark.
Fewer false-positive refusals. Anthropic says its cybersecurity safeguards now intervene about 60% less often on benign work, and Fable 5.1 can be used to identify software vulnerabilities (though not to build exploits — penetration testing and exploit generation still get redirected to Opus models). Benign biology and medical questions trigger safeguards 85% less often than at Fable 5's launch.
Fable 5.1 vs Mythos 5.1
They are the same model with different safeguards. Fable 5.1 is the generally available version. Mythos 5.1 removes certain cybersecurity and life-sciences restrictions and is only available to vetted organisations through Anthropic's trusted access programs — a Cyber Verification Program for defensive security work and a Life Sciences Verification Program run in partnership with the US government.
For almost everyone reading this, "Claude Fable 5.1" is the model you can actually use, and the capability difference only shows up on the restricted task types. On Anthropic's own coding benchmark the gap is a few points, and they expect the new safeguards to narrow it further.
Pricing, and what it means for subscribers
On the API: $10 per million input tokens, $50 per million output — identical to Fable 5 — with cache reads at $0.25 per million. Anthropic's own estimate is ~25% lower total cost on typical workloads and up to ~45% on heavily agentic ones.
On a Claude subscription, the calculus is about your usage allowance rather than dollars. Fable remains the heaviest consumer of the lineup, so running it as your daily default is the fastest route to hitting your usage limit. The effort setting softens this — Medium effort buys Fable-5-class results for less — but for routine writing and everyday coding, Sonnet 5 still does the job without touching your allowance anywhere near as hard.
Who should actually use it
Switch for: long-running agentic sessions, debugging that has stumped other models, multi-service refactors, research-style problems where the reasoning is genuinely the hard part. Anthropic's early-access partners consistently describe it as the model for unattended, multi-hour work.
Don't switch for: everyday chat, routine drafting, quick questions. Nothing about this release changes the default-to-Sonnet rule — it just raises the ceiling for the tasks that escalate past it.
It also supports the 1M-token context window, alongside Fable 5, Sonnet 5, Opus 4.6 and later, and Sonnet 4.6 — see Claude Code's context window for how that works in practice.
Common questions
What is Claude Fable 5.1?
How much does Fable 5.1 cost?
What is Claude Mythos 5.1?
Is Fable 5.1 better than Opus 5?
Fable 5.1 is live on Claude now
It's available on paid plans today — but try your hard tasks on it specifically. For everyday work, Sonnet 5 still stretches your allowance much further.