On June 9, 2026, Anthropic released Claude Fable 5 — the first publicly available model in a new "Mythos-class" tier that sits above the Opus line. It is, by every available measure, the most capable AI model currently on the market. But it also ships with the most aggressive safety scaffolding Anthropic has ever put on a general release, making it one of the most interesting launches in the industry's short history.
What Is Claude Fable 5?
Fable 5 is the public-facing sibling of two models built on the same weights:
- Claude Fable 5 — available through the API (model ID:
claude-fable-5), the Claude app, and Amazon Bedrock. This is the one developers and businesses can use. - Claude Mythos 5 — the same underlying model with certain safeguards lifted, available only through Anthropic's Project Glasswing program for vetted cybersecurity researchers and infrastructure providers.
The two-model approach is Anthropic's answer to a hard problem: how do you release a model capable enough to assist in cyberattacks without making those capabilities freely available? Fable 5 handles sensitive queries by falling back to Claude Opus 4.8 rather than refusing entirely — a design that Anthropic says triggers in less than 5% of sessions but that some early users have found overly conservative.
Benchmarks: A Clear Step Up
Anthropic has not yet published a full system card with official numbers, so all scores below come from independent third-party testing. Treat them as early signals rather than final figures.
| Benchmark | Claude Fable 5 | Claude Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro |
|---|---|---|---|---|
| SWE-Bench Pro (agentic coding) | ~80.3% | ~69.2% | ~58.6% | ~54.2% |
| Terminal-Bench 2.1 | ~88.0% | — | ~83.4% | — |
| Humanity's Last Exam (no tools) | ~59.0% | — | ~52.2% | — |
| Humanity's Last Exam (with tools) | ~64.5% | — | — | — |
| FrontierCode Diamond | ~29.3% | ~14% | — | — |
The standout is agentic coding. Independent SWE-Bench Pro testing puts Fable 5 around 80% — roughly 11 points above Opus 4.8 and over 20 points above GPT-5.5. During early testing, Stripe reported that Fable 5 compressed months of engineering into days, completing a codebase-wide migration across a 50-million-line Ruby codebase in a single day.
Where Fable 5 Shines
Software engineering. Beyond raw benchmark scores, the practical difference is autonomy. Fable 5 can run complex coding tasks for extended periods with minimal human intervention. On Cognition's FrontierCode evaluation, which tests whether models can pass difficult tasks while meeting production code quality standards, Fable 5 scores highest among frontier models even at medium effort levels.
Knowledge work. On Hebbia's Finance Benchmark for senior-level reasoning, Fable 5 posts the highest score of any model, with gains in document-based reasoning, chart and table interpretation, and problem solving. IMC reported that it aced trading-analysis evaluations nearly across the board.
Vision. Fable 5 is the new state-of-the-art for vision tasks. It can extract precise numbers from scientific figures and reconstruct web app source code from screenshots. In a fun demonstration, earlier Claude models needed a complex helper harness to play Pokémon FireRed — Fable 5 beat the game using raw screenshots alone.
Memory and long-context. When Anthropic tested Fable 5 on the deck-building game Slay the Spire, giving it access to persistent file-based memory improved its performance three times more than the same setup did for Opus 4.8. The model uses its own notes to improve outputs over long-running tasks.
Life sciences. Mythos 5 (the unrestricted version) has already been used by Anthropic's internal protein design experts to accelerate drug discovery workflows. The company cites novel hypothesis generation and faster therapeutic development as early results.
Pricing and Availability
Fable 5 is priced at exactly twice Opus 4.8:
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Claude Fable 5 | $10.00 | $50.00 |
| Claude Opus 4.8 | $5.00 | $25.00 |
Prompt caching and batch rates are available at a discount. During the introductory period (June 9–22, 2026), Fable 5 is included free for Claude Pro, Max, Team, and Enterprise users. After June 23, usage draws on standard credits and billing.
The model is available on the Claude API, the Claude app, and Amazon Bedrock.
What This Means
Fable 5 represents a shift in how AI companies handle the capability-safety tradeoff. Rather than holding back their most powerful models, Anthropic is releasing them with guardrails that route sensitive queries to a less capable model. It is not a perfect solution — the fallback to Opus 4.8 can feel jarring, and some legitimate queries get caught — but it is a pragmatic approach that gets frontier capabilities into developers' hands faster.
The "Mythos-class" branding also signals that Anthropic sees Opus as no longer the ceiling. If Fable 5 is the first Mythos model, more will follow. For developers choosing between API providers, the calculus just changed: Anthropic now offers a model that leads on the benchmarks that matter most for production workloads, at a price that undercuts the competition on a per-task basis despite the higher per-token cost.
The question for the next few months is how quickly Anthropic can reduce false positives in the safety fallback, and whether the Glasswing program gives Mythos 5 enough real-world testing to justify broader access. For now, Fable 5 is the model to beat.
Looking for AI models for your own projects? Browse the Q4KM model directory to compare thousands of models across benchmarks, hardware requirements, and licensing.