Claude Fable 5.1 Pricing, Benchmarks, and What Changed From Fable 5

Claude Fable 5.1 Pricing, Benchmarks, and What Changed From Fable 5

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, the first update to Fable 5 since it launched June 9, 2026. The headline number isn't the base price, which didn't move. It's cache reads, cut 75% from $1.00 to $0.25 per million tokens, which is where Anthropic says most of its claimed 25 to 45% cost savings actually come from. Here's what changed, what the benchmarks show, and where Fable 5.1 fits next to Opus 5.

What is Claude Fable 5.1?

Claude Fable 5.1 is Anthropic's first update to the Fable line since Fable 5's June 9, 2026 launch. It ships alongside Claude Mythos 5.1, a restricted-access version of the same underlying model with reduced safety filters. Fable 5.1 is generally available with standard safeguards. Mythos 5.1 is gated to vetted organizations through two channels: cybersecurity groups vetted under Anthropic's Cyber Verification Program, and life-sciences organizations vetted under the Life Sciences Verification Program, a partnership with the US government.

That access split matters for anyone comparing benchmark numbers between the two models, which we get to below. If you're setting up access to either model, our guide on accessing Claude Fable 5 and Mythos 5 covers the account and eligibility requirements that still apply to the 5.1 versions.

What changed in Claude Fable 5.1's pricing?

Base input and output pricing is unchanged from Fable 5. The change is entirely in caching.

(Source: platform.claude.com model overview, anthropic.com Fable and Mythos 5.1 announcement)

We covered Fable 5's original $10/$50 pricing in detail in our Fable 5 pricing breakdown. The rate card there for input, output, and batch pricing still applies to 5.1. Only the cache-read line changed.

How much cheaper is Fable 5.1 in practice?

Anthropic's own estimate: roughly 25% cheaper for typical workloads, and up to roughly 45% cheaper for highly agentic workloads, compared to Fable 5 at the same base prices. Both figures come entirely from the cache-read cut, not from any reduction in input or output pricing.

That framing matters if you're budgeting off list price alone. A workload that rarely hits the prompt cache (short, one-off requests) will see close to 0% savings, since input and output rates didn't change. A workload with a long, stable system prompt or repeated context, the kind of setup agentic coding tools and multi-turn agents lean on, hits the cache constantly, which is why Anthropic's high end of the savings range is tied to "highly agentic" use. If your workload doesn't cache much, don't expect the discount.

How does Fable 5.1 perform on benchmarks vs Fable 5 and Opus 5?

Anthropic published its own comparison table at launch, covering Fable 5.1, Mythos 5.1, Fable 5, Opus 5, and GPT-5.6 Sol.

(Source: anthropic.com Fable and Mythos 5.1 comparison table)

GDPval-AA v2 is a raw score, not a percentage, so read that row as a relative ranking rather than a pass rate. On every benchmark Anthropic listed, Fable 5.1 beats both Fable 5 and Opus 5, except one.

Why does Mythos 5.1 beat Fable 5.1 on Terminal-Bench 4.0?

On Terminal-Bench 4.0, the top score belongs to Mythos 5.1 at 60.9%, not Fable 5.1's 55.8%. That's the one benchmark in Anthropic's table where the restricted-access sibling model outscores the generally available one. Anthropic's table also lists Mythos 5 (the prior generation) at 42.0% on the same benchmark, well below both 5.1 versions.

Since Fable 5.1 and Mythos 5.1 share the same underlying model and differ only in safety filtering, the gap suggests the reduced safeguards on Mythos 5.1 change its behavior on agentic coding tasks specifically, at least on this benchmark. Anthropic hasn't published a breakdown of why. Because Mythos 5.1 access is restricted to vetted cybersecurity and life-sciences organizations, most teams evaluating Fable 5.1 for general use won't have a way to reproduce that number themselves.

Should you use Fable 5.1 or Opus 5?

Anthropic's own documentation recommends starting with Claude Opus 5 for most workloads, and reaching for Fable 5.1 specifically for "demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5 at higher effort still fall short." That's Anthropic's stated positioning, not a marketing claim we're adding on top of it.

In practice, that means Fable 5.1 isn't the default pick. If Opus 5 already clears your eval bar, there's no benchmark result here that argues for switching. The published numbers above are the case for Fable 5.1 when Opus 5 doesn't clear that bar, particularly on Terminal-Bench-Science 0.1 and OSWorld 2.0, where the gap to Opus 5 is largest. For a closer side-by-side of Fable-generation pricing and specs against Anthropic's other current model, see our Fable 5 vs Opus 4.8 comparison. If neither Anthropic model fits your workload or budget, our Fable 5 alternatives roundup covers other options.

What did Anthropic show off at launch?

Alongside the benchmarks, Anthropic cited a handful of applied results:

• Protein design: work across 12 targets hit "nearly 50%" hit rate, with binding affinities Anthropic described as "10 times higher" than competing approaches.
• Venus mapping: surface-detail resolution improved to "two to three kilometers," up from a prior "10 to 20 kilometers."
• Computational biology: GPU speedups of "up to 2.5 times," with 30-60% cost savings on genome-wide analyses.

None of these are independently verified benchmarks. They're Anthropic's own reported results from its launch page.

One named third-party example did surface in press coverage: 9to5Mac reported that Millennium, an investment firm, tested Fable 5.1 and had it identify the cause of a rare system crash that engineers and other models had failed to resolve over several years (source: 9to5mac.com, Zac Hall, September 1, 2026). That's a single anecdote from one firm, not a benchmark, but it's the clearest real-world example attached to this launch so far.

Fable 5.1 specs at a glance

(Source: platform.claude.com model overview)

Fable 5.1 sits in our full model comparison hub alongside every other current Claude, GPT, and Gemini release, if you want the same specs next to competing models.

FAQ

What is Claude Fable 5.1?

Claude Fable 5.1 is Anthropic's first update to Claude Fable 5, released September 1, 2026. It's the same underlying model as its restricted-access sibling, Claude Mythos 5.1, but with standard safety filters instead of Mythos's reduced safeguards for vetted cybersecurity and life-sciences organizations.

How much does Claude Fable 5.1 cost?

Input tokens are $10 per million and output tokens are $50 per million, unchanged from Fable 5. The actual price change is in caching: cache reads dropped 75%, from $1.00 to $0.25 per million tokens, plus new cache-write rates of $12.50/MTok (5-minute) and $20/MTok (1-hour). Batch API stays at 50% off base rates.

Is Claude Fable 5.1 cheaper than Fable 5?

Only for workloads that use prompt caching. Anthropic estimates roughly 25% cheaper for typical workloads and up to roughly 45% cheaper for highly agentic workloads, both driven entirely by the cache-read discount. A workload with little or no caching sees close to no savings.

What's the difference between Claude Fable 5.1 and Claude Mythos 5.1?

They're the same underlying model with different safety filtering. Fable 5.1 is generally available with standard safeguards. Mythos 5.1 is restricted to vetted cybersecurity organizations (via the Cyber Verification Program) and life-sciences organizations (via the Life Sciences Verification Program), a partnership with the US government.

Should I use Claude Fable 5.1 or Claude Opus 5?

Anthropic recommends starting with Opus 5 for most workloads, and moving to Fable 5.1 for demanding reasoning and long-horizon agentic work, or when Opus 5 at higher effort still falls short of your evals. Fable 5.1 beats Opus 5 on every benchmark Anthropic published except Terminal-Bench 4.0, where Mythos 5.1, not Fable 5.1, posts the top score.

What is Claude Fable 5.1's context window?

1,048,576 tokens (1M), with a maximum output of 128,000 tokens. The API identifier is claude-fable-5-1, the reliable knowledge cutoff is June 2026, and Anthropic has committed not to retire the model before September 1, 2027.

Where is Claude Fable 5.1 available?

Claude Code, the Claude Platform (Enterprise tier), the Anthropic API, AWS Bedrock, Google Cloud, and Microsoft Foundry.

---

Sources: Anthropic, Claude Fable and Mythos 5.1 announcement, Anthropic model overview docs, 9to5Mac, Zac Hall