Claude Fable 5.1 Release Date: What We Know So Far

Anthropic launched Claude Fable 5.1 on September 1, 2026. That is the release date, and it ends weeks of leak-driven speculation. The model is generally available today across the Claude API, Amazon Web Services, Google Cloud, and Microsoft Azure, callable as claude-fable-5-1. Alongside it, Anthropic released Claude Mythos 5.1, the same underlying model with lighter safeguards, restricted to vetted organizations through trusted-access programs.

The leaks got the big things right. Pricing held at Fable 5's rates, and the focus landed on long-horizon agentic work. For the day-one announcement, read our launch coverage. This page is the standing answer to the release-date question: the confirmed date and rollout, the pricing change that actually matters, the benchmark results, and how Mythos 5.1 fits in.

When Claude Fable 5.1 released and where you can use it

The Claude Fable 5.1 release date is September 1, 2026, and unlike a phased enterprise preview, this one went straight to general availability. Developers can call claude-fable-5-1 on the Claude API and through Amazon Bedrock, Google Cloud, and Microsoft Foundry the same day. Anthropic frames Fable 5.1 as its most advanced model for coding, knowledge work, and long-running problem-solving, positioned above the Opus, Sonnet, and Haiku tiers.

The release confirms what the earlier leaks pointed to. Reports in late July had flagged an August-to-September window with unchanged pricing, and both held up. The one detail the leaks understated is the cache-read price cut, which turns out to be the most consequential change in the launch.

Fable 5.1 defaults to different effort levels depending on where you use it: High in Claude Code, and Medium in Claude Cowork and on Claude.ai. That matters for cost, because Anthropic says Fable 5.1 at Low or Medium effort matches or beats Fable 5's results at a much lower price.

What Claude Fable 5.1 costs

Claude Fable 5.1 keeps Fable 5's headline pricing: $10 per million input tokens and $50 per million output tokens. The real change is the cache-read rate, which drops 75% to $0.25 per million tokens, down from $1. That single cut reshapes the effective cost of running the model on repeated-context work.

Anthropic estimates the saving at around 25% for typical workloads and up to roughly 45% for complex coding and highly agentic tasks, where cache reads make up most of the bill. Those figures come from Anthropic's own four-week usage measurement in August 2026, so treat them as vendor estimates tied to a specific workload mix rather than a guaranteed discount.

Pricing as of September 2026. Source: Anthropic.

The cache cut produces an unusual pricing profile. Fable 5.1's base input and output are twice Opus 5's, yet its cached input is cheaper than Opus 5's cache reads, and only about 25% above Sonnet 5's despite Fable's base input price being five times higher. For context-heavy, tool-heavy workloads that reuse a lot of the same input, that math changes which tier makes sense.

Claude Fable 5.1 pricing on the Claude API - input, output, and cache-read rates

The benchmarks are strong, with caveats worth reading

Claude Fable 5.1 posts clear gains over Fable 5 across coding, science, and knowledge work, though every figure here is Anthropic-reported rather than independently verified. Anthropic evaluated Fable 5.1 with its production safeguards enabled, and it flags that on some tasks those safeguards intervened, which can lower the reported score. All numbers below carry that vendor-reported caveat, and our full benchmark breakdown covers how they hold up against independent testers.

On agentic terminal coding, Fable 5.1 scores 55.8% on Terminal-Bench 4.0, up from 42.0% for Fable 5, and ahead of Opus 5 at 52.3% and GPT-5.6 Sol at 37.3%. On agentic scientific research, it reaches 52.6% on Terminal-Bench-Science 0.1, more than double Fable 5's reproduced 24.7%. Anthropic notes a standard error of roughly 3.5 to 4.5 points on that benchmark, so treat close gaps as noise rather than clean separation.

Selected Anthropic-reported benchmark results. All scores are vendor-reported at max effort

The comparison against Opus 5 is the one that matters for buyers. Fable 5.1 opens a measurable lead over Opus 5 on the hardest agentic and scientific tasks, which is the gap the Fable tier needs to justify its price premium. On more general knowledge work, the two sit closer together, so the case for paying the Fable rate rests on long-horizon, multi-step autonomy rather than everyday coding. Our Fable 5.1 vs Opus 5 guide breaks down which to pick per project.

One benchmark to read carefully is OSWorld 2.0. Anthropic reports 77.9% for Fable 5.1 on a partial-credit basis, but notes the task files differ from earlier OSWorld releases, so the number is not directly comparable to previously published scores. That is a disclosure worth respecting rather than a figure to quote flatly against rivals.

How Mythos 5.1 fits in

Claude Mythos 5.1 is the same model as Fable 5.1, with more permissive safeguards, and it is not generally available. Anthropic restricts it to vetted individuals and organizations through two trusted-access programs: the Cyber Verification Program for defensive security work, and the Life Sciences Verification Program for professional research, the latter developed in partnership with the US government. For now, access is limited to a set of US organizations, with expansion planned.

The split exists because Fable 5.1's general-use safeguards redirect certain dual-use tasks to Anthropic's Opus models. Penetration testing, exploit generation, and binary-based vulnerability scanning still route away from Fable 5.1 for standard users. Mythos 5.1 is how Anthropic gives cleared cyberdefenders and life scientists access to those capabilities under safeguards designed for professional work.

Anthropic did loosen one restriction for everyone. Fable 5.1 can now be used to identify software vulnerabilities, the kind of defensive work that improves software security, even though developing exploits from them remains gated. The company says Claude Code users can expect around 60% fewer interventions per session from its cyber safeguards compared with Fable 5.

What changed from Fable 5

The upgrade from Fable 5 to Fable 5.1 is mostly about efficiency and fewer false positives, not a headline capability leap. The cache-read cut is the biggest practical change, followed by safeguard tuning that reduces how often benign requests get flagged. Our Fable 5.1 vs Fable 5 comparison covers the full list, but the two figures that stand out are these: Anthropic reports its newest cybersecurity safeguards block 60% fewer false positives, and its biology safeguards fire 85% less often on elementary biology and medical questions.

Anthropic also cites qualitative gains its early partners reported: better writing quality, steadier performance on long unattended runs, and more readable output over multi-step tasks. One example the company highlighted is a test at the investment firm Millennium, where Fable 5.1 traced the cause of a rare crash that had gone unexplained for years and that no other model had cracked. Those are vendor-supplied anecdotes, useful as signal but not as proof.

Developers upgrading should note the change is not purely additive. New API accounts can no longer manually edit Claude's prior thinking in a multi-turn conversation, an anti-distillation measure, and there are documented behavior changes around tool choice and thinking blocks. Existing integrations are worth testing before a production switch.

Stay on top of the latest AI model news

Claude Fable 5.1 is one launch in a release cycle that now moves week to week, and its release date was a moving target until the day it shipped. The next flagship, from Anthropic or a rival, will follow the same arc: leaks, prediction-market odds, and community sightings, then a launch that confirms or corrects all of it.

For the running story, Emergent News tracks it as it happens, from early rumors and release-date speculation on upcoming models to the confirmed details when a new model actually ships. It is the fastest way to know what is real, what is still unconfirmed, and what just launched.