Skip to content
News Model release

Anthropic launches Claude Fable 5 and Mythos 5, first Mythos-class models for general use

· by Pondero Newsdesk

The short version

Anthropic released Claude Fable 5 and Claude Mythos 5 on June 9, 2026, making a Mythos-class model available for general use for the first time and pricing both at $10 per million input tokens and $50 per million output tokens.

Anthropic launches Claude Fable 5 and Mythos 5, first Mythos-class models for general use

Anthropic released Claude Fable 5 and Claude Mythos 5 on June 9, 2026, giving developers and enterprises access to a Mythos-class model for the first time. Both models are priced at $10 per million input tokens and $50 per million output tokens, per Anthropic's announcement.

What

The two models share the same underlying architecture. The difference is safeguards. Fable 5 includes classifiers that intercept queries in cybersecurity, biology, chemistry, and distillation detection, routing those requests to Claude Opus 4.8 instead of responding directly. Mythos 5 lifts those restrictions for an approved set of partners through Project Glasswing.

Anthropic describes Fable 5 as exceeding "the capabilities of any model we've ever made generally available," calling it "state-of-the-art on nearly all tested benchmarks of AI capability." In early testing, Stripe reported that Fable 5 compressed a codebase-wide Ruby migration of a 50-million-line repository into a single day, a task that would otherwise have required a full engineering team more than two months to complete by hand, per Anthropic's announcement. Hebbia reported the highest score of any model on its Finance Benchmark for senior-level reasoning. Cursor CEO Michael Truell called Fable 5 "the state of the art model on CursorBench," adding that it "opened up a class of long-horizon problems that were out of reach for earlier models."

Safeguards trigger in fewer than 5% of Fable 5 sessions, meaning the vast majority of users receive Fable-class responses with no fallback to Opus 4.8. The conservative tuning means some benign requests will trip the classifiers; Anthropic says it plans to reduce false positives after launch. A data retention policy accompanies the release: Anthropic will require 30-day retention for all Mythos-class model traffic, which will not be used for training.

On the pricing side, $10/$50 per million tokens is described as "less than half the price of Claude Mythos Preview," making Fable 5 the most cost-efficient way to access Mythos-class capability, per Anthropic. Subscription plan holders on Pro, Max, Team, and seat-based Enterprise get Fable 5 included at no extra cost through June 22. Starting June 23, subscription users will need usage credits to access Fable 5; Anthropic said it intends to restore Fable 5 to standard subscription plans once capacity allows.

Why it matters

Fable 5 closes the gap between what frontier research teams can access and what general developers can access. Before this release, Mythos-class capability was restricted to a small number of cyber defenders and critical infrastructure providers through Project Glasswing. Fable 5 puts nearly the same model into the standard API, Amazon Bedrock, Vertex AI, Microsoft Foundry, and GitHub Copilot. Developers building agents, code tools, or analytical pipelines can now call these capabilities at API rates with no special access request.

The safeguard architecture itself has implications beyond this launch. Anthropic's framing positions the fallback-to-Opus-4.8 design as a repeatable pattern for releasing frontier-class capabilities safely. Rather than holding back the full model indefinitely, the company ships the model with classifiers in front of sensitive domains and tuned conservatively, then narrows the coverage over time. That approach, if it holds, means future capability releases from Anthropic could move faster than pre-Fable precedent suggests.

For GitHub Copilot users specifically, Fable 5 is now selectable in the model picker across all Copilot tiers, per the GitHub Changelog. GitHub's CPO Mario Rodriguez said in early testing Fable 5 "took on complex, long-horizon coding tasks with a level of autonomy and reliability that exceeded previous benchmarks." If you are evaluating whether Fable 5 changes your team's Copilot setup, GitHub Copilot plans are available here.

Context and reactions

Claude Mythos 5 is initially available only to existing Project Glasswing partners as an upgrade to Claude Mythos Preview, deployed in collaboration with the US government. Anthropic described Mythos 5 as having "the strongest cybersecurity capabilities of any model in the world," per the same announcement. Plans include a broader trusted access program for both cybersecurity organizations and, separately, a biology trusted access program for life sciences researchers.

Augment Code shipped Fable 5 integration on the same day as the launch. Cursor issued a statement pointing to CursorBench results. Several other development platforms are expected to add Fable 5 to their model pickers in the days following the announcement.

The launch comes roughly eight weeks after Anthropic introduced Claude Mythos Preview through Project Glasswing in April. At that time, Anthropic wrote that it hoped to eventually release Mythos-level capabilities broadly "so long as we had developed new safeguards that were strong enough to reliably prevent misuse." Fable 5 is the company's declaration that those safeguards now meet that bar.

TechCrunch noted that the release comes days after Anthropic published warnings about the risks of increasingly capable AI systems, per TechCrunch's June 9 report. The company addressed that tension directly in its announcement, framing the safeguard classifiers as the mechanism that makes a general release responsible at this capability level.

On the AWS side, Fable 5 on Amazon Bedrock is immediately available with the same safeguards as the API version, per AWS's announcement. AWS customers with existing Bedrock API access do not need to request new permissions. Vertex AI and Microsoft Foundry availability shipped the same day.

What to watch next

Anthropic said it intends to narrow the biology and chemistry classifiers in coming weeks, with a trusted access program for life sciences researchers arriving shortly after launch. Whether the fallback rate declines as classifiers improve is a concrete near-term signal of how aggressively Anthropic can reduce false positives without increasing actual risk. Cursor's model picker update will be the first third-party indicator of where Fable 5 lands in real developer workflow adoption. Gemini's expected June launch is the next direct benchmark comparison point.

Sources