Claude Fable 5 & Mythos 5: Anthropic Splits the Frontier Into Two Products
On June 9, 2026, Anthropic shipped a single frontier model as two products: Claude Fable 5 for everyone, with safeguards, and Claude Mythos 5 for vetted partners, with some safeguards lifted. Here is what is real, what is restricted, and what it changes for teams building with AI.
TL;DR
- One model, two products. Fable 5 and Mythos 5 share the same underlying model. Fable 5 is generally available with safety classifiers; Mythos 5 is the same model with some safeguards lifted, restricted to vetted partners.
- It leads the board on real work. Fable 5 tops agentic coding (SWE-Bench Pro 80.3% vs 69.2% for Opus 4.8), knowledge work, tool use, legal, and health.
- Read the asterisks. The dazzling cybersecurity and biology numbers belong to Mythos 5 — the model you cannot buy. Fable 5 falls back to Opus 4.8 on those exact topics.
- It is a premium tier. $10 / $50 per million tokens — double Opus 4.8. Reach for it on the hardest long-horizon work; keep Opus 4.8 as the default for routine tasks.
Executive Summary
On June 9, 2026, Anthropic did something it had not done before: it shipped a single frontier model as two distinct products.[1] Claude Fable 5 (API model ID claude-fable-5) is the new generally available flagship, wrapped in safety classifiers that hand a narrow set of sensitive queries off to Claude Opus 4.8. Claude Mythos 5 is the same underlying model with those safeguards lifted in some areas — and it is not generally available. Access is restricted to vetted partners in Anthropic's Project Glasswing program, working on cyberdefense and critical infrastructure.
Both models belong to a new tier Anthropic calls "Mythos-class," which sits above the Opus class. The crucial detail: the underlying model is the same in both cases. The difference between them lives entirely in the safeguards layered on top. For the overwhelming majority of business work — coding, analysis, content, research, agentic workflows — the model you can actually deploy, Fable 5, runs at full strength.
One Model, Two Products
Start with the part that is easy to miss in the benchmark excitement: Fable 5 and Mythos 5 are not two different models. They are the same model, packaged two ways. Anthropic's framing is that the raw model is capable enough in areas like cybersecurity and biology that releasing it openly would be irresponsible. So the company split the release — Fable 5 for everyone with guardrails, Mythos 5 for a small set of approved organizations with some of those guardrails removed.
Claude Fable 5
API ID claude-fable-5. Live on the Claude API, claude.ai (Pro / Max / Team / Enterprise), and AWS, Google Cloud & Microsoft Foundry.
Safety classifiers route cybersecurity, biology/chemistry, and model-distillation queries to Opus 4.8.
Claude Mythos 5
No public API ID. Deployed via Project Glasswing to cyberdefenders and infrastructure providers, with trusted-access programs planned.
The same model with safeguards lifted in some areas — described as the strongest cybersecurity model in the world.
Two products from one base model. The line between them is governance, not capability.
When a query lands on a safeguarded topic, Fable 5 quietly routes the response to Claude Opus 4.8 instead of refusing it. Anthropic reports these classifiers trigger on fewer than 5% of sessions on average,[1] so for routine work the handoff is invisible. That detail is the whole reason the company felt able to release a Mythos-class model to the public at all.
Benchmark Analysis — The Five-Model Matrix
Anthropic published a comparison across Claude Fable 5 / Mythos 5, Claude Mythos Preview, Claude Opus 4.8, GPT-5.5, and Gemini 3.1 Pro.[1] Two methodology points are load-bearing. First, the table shows the higher of the Fable 5 and Mythos 5 scores, which are within one to three percentage points of each other on most benchmarks. Second, starred (*) rows are where the two diverge more — because Fable 5's blocking safeguards for cybersecurity and biology pull its score down toward Opus 4.8. On those rows, the number you see is Mythos 5; the number a Fable 5 deployment delivers is lower.
Figure 1 — SWE-Bench Pro (Agentic Coding), Five Models
Fable 5 leads agentic coding by a wide margin — +11.1 points over Opus 4.8. Source: Anthropic benchmark table, Jun 9, 2026.
The full published matrix. Starred (*) rows report Mythos 5; a Fable 5 deployment performs closer to Opus 4.8 on those:
| Benchmark | Fable 5 / Mythos 5 | Mythos Preview | Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro |
|---|---|---|---|---|---|
| Agentic coding — SWE-Bench Pro | 80.3% | 77.8% | 69.2% | 58.6% | 54.2% |
| Agentic coding — FrontierCode (Diamond, xhigh) | 29.3% | — | 13.4% | 5.7% | — |
| Agentic coding — Terminal-Bench 2.1 | 88.0%* | — | 82.7% | 83.4% | 70.7% |
| Knowledge work — GDPval-AA (ELO) | 1932 | — | 1890 | 1769 | 1314 |
| Knowledge work (vision) — GDP.pdf, no tools | 29.8% | — | 22.5% | 24.9% | 16.7% |
| Spatial reasoning — Blueprint-Bench 2 | 38.6% | — | 14.5% | 36.2% | 26.5% |
| Tool use — AutomationBench | 17.4% | — | 15.5% | 12.9% | 9.6% |
| Computer use — OSWorld-Verified | 85.0% | 85.4% | 83.4% | 78.7% | 76.2% |
| Legal — Legal Agent Benchmark | 13.3% | — | 10.4% | 2.1% | 0.0% |
| Reasoning — Humanity's Last Exam (no tools) | 59.0%* | 56.8% | 49.8% | 41.4% | 44.4% |
| Reasoning — Humanity's Last Exam (with tools) | 64.5%* | 64.7% | 57.9% | 52.2% | 51.4% |
| Biology — BioMysteryBench (hard) | 46.1%* | 29.6% | 40.0% | — | — |
| Cybersecurity — ExploitBench (Cap%) | 78.0%* | 69.0% | 40.0% | 34.0% | — |
| Health — HealthBench Professional | 66.0%* | 64.7% | 56.9% | 51.8% | — |
The clean reads: on agentic coding, knowledge work, spatial reasoning, tool use, legal, and health, the new model shows the same shape — a clear lead over Opus 4.8 and the broader field. The one row it does not lead is computer use, where Mythos Preview edges it 85.4% to 85.0% — a tie within noise.
The Safeguard Architecture
The split is built on safety classifiers — separate systems that watch the conversation and detect attempts at misuse. When a classifier fires on Fable 5, the response is not refused outright; it is handed to Claude Opus 4.8, the next-most-capable model, which answers in Fable 5's place. Because the classifiers trigger on under 5% of sessions on average, the handoff is invisible for the work most teams do.
Most traffic passes the gateway at full capability; a small fraction is rerouted to an Opus 4.8 fallback.
Three domains are covered by the safeguards. The flow below shows how a query is handled:
Figure 2 — How Fable 5 Routes a Query
Cybersecurity, biology/chemistry, and distillation queries fall back to Opus 4.8. Everything else runs on Fable 5.
- Cybersecurity — exploitation, offensive cyber tasks, and agentic hacking. This is why ExploitBench shows 78.0% for Mythos 5, but Fable 5 made 0% progress on offensive cyber tasks in blocking mode.
- Biology & chemistry — currently a broad net that Anthropic acknowledges is overly wide, with narrowing planned so legitimate biomedical work is not caught. Watch this if your work touches health, pharma, or life sciences.
- Distillation — extraction attacks designed to siphon the model's behavior to train a competing model.
On robustness: Anthropic ran more than 1,000 hours of external bug-bounty red-teaming and reports no universal jailbreak was found, though the UK AI Safety Institute made early progress toward one within an initial testing window. Mythos-class traffic carries a 30-day retention requirement, is not used for training, and logs all human access.[1]
Pricing, Access & Rollout
Fable 5 runs at $10 per million input tokens and $50 per million output tokens — double Opus 4.8's $5 / $25 rate card. A 90% discount applies to input via prompt caching, and US-only inference is available at a 1.1× multiplier. Unlike the same-price Opus 4.7-to-4.8 jump, this one has a genuine cost decision attached.
| Dimension | Claude Fable 5 | Claude Opus 4.8 |
|---|---|---|
| Input price (per 1M tokens) | $10 | $5 |
| Output price (per 1M tokens) | $50 | $25 |
| Availability | General — API, claude.ai, AWS, GCP, Foundry | General |
| SWE-Bench Pro | 80.3% | 69.2% |
| Best for | Hardest, long-horizon coding & knowledge work | Routine, high-volume, latency-sensitive work |
The rollout you need to plan around. On subscription plans, Fable 5 is included at no extra cost from June 9 through June 22. From June 23, using it on a subscription plan draws on usage credits; API and cloud usage is billed per token throughout. If your team is building a Fable 5 evaluation on claude.ai, run it inside the free window and budget for credits afterward.
When to Use Fable 5 (and When Opus 4.8 Is Still the Call)
Fable 5 is built for ambitious, long-running, asynchronous tasks that previous models could not sustain: agents that work for hours or days, plan across stages, and delegate to sub-agents. At the highest effort it reflects on and validates its own output. Its vision is strong on the messy real-world inputs that matter in business — diagrams, charts, and tables nested inside files and PDFs.
| Reach for Claude Fable 5 when… | Stay on Claude Opus 4.8 when… |
|---|---|
| Migrating a framework across thousands of files | Classifying, routing, or summarizing at high volume |
| Running a multi-day research-and-build task | Powering interactive chat where latency matters |
| A missed detail costs more than the token bill | Cost-sensitive, routine agentic tasks |
| The work benefits from sustained autonomy + self-verification | The workload lives near cybersecurity or biology (Fable 5 falls back to Opus 4.8 anyway) |
The disciplined approach: run a representative sample of your real tasks on both tiers, measure the quality difference and the token spend, and route by task rather than standardizing on one model.
Read the Asterisks Before You Quote the Numbers
The single most important caveat for anyone evaluating this release: the starred cyber and biology figures belong to the restricted model. A Fable 5 deployment performs closer to Opus 4.8 on those exact tasks. The clearest case is ExploitBench — 78.0% for the model behind the gate, while Fable 5 made 0% progress on offensive cyber tasks in blocking mode. If a vendor, a slide, or a comparison cites the 78.0% number as if it were the model you can buy, it is overstating what you will actually receive.
- The price step-up is real. At $10 / $50, Fable 5 is twice Opus 4.8. Model your task distribution before standardizing on it.
- Biology safeguards are broad for now. Until they narrow, legitimate biomedical, pharma, and life-sciences queries may get the Opus 4.8 fallback. If that is your domain, test before you commit.
- Some specs are not public. Anthropic's announcement did not state Fable 5's context-window size or maximum output tokens. Verify these against the model documentation before architecting around a specific context length.
Frequently Asked Questions
What is the difference between Claude Fable 5 and Claude Mythos 5?
They are the same underlying model, packaged two ways. Fable 5 (claude-fable-5) is generally available with safety classifiers that route cybersecurity, biology/chemistry, and distillation queries to Opus 4.8. Mythos 5 is that same model with those safeguards lifted in some areas, restricted to vetted Project Glasswing partners.
How much does Claude Fable 5 cost?
$10 per million input tokens and $50 per million output tokens — double Opus 4.8's $5 / $25. A 90% prompt-caching discount applies to input, and US-only inference is offered at a 1.1× multiplier.
How does Fable 5 compare to Opus 4.8?
It leads on agentic coding (SWE-Bench Pro 80.3% vs 69.2%), knowledge work (GDPval-AA 1932 vs 1890), spatial reasoning, tool use, legal, and health. It is built for long-horizon, self-validating work. Opus 4.8 stays the sensible default for routine, cost- or latency-sensitive tasks.
Is Fable 5 safe to use for business work?
Yes. The safety classifiers trigger on fewer than 5% of sessions, and for the work most businesses do — coding, analysis, content, research, agentic delivery — the routing never fires and Fable 5 runs at full strength.
Conclusion
The headline is the split. Anthropic decided its newest model was capable enough in cybersecurity and biology that releasing it openly would be irresponsible, so it shipped two products from one model: Fable 5 for general use with safeguards, and Mythos 5 for vetted partners with some safeguards lifted. That governance choice is as significant as any benchmark on the page.
For teams, the practical read is clear. Fable 5 leads the field on the capabilities businesses actually use, and it is built for the long-horizon, self-validating, sub-agent work that is hard to get from a cheaper model. Keep two things straight: the dazzling cyber and biology numbers belong to a model you cannot buy, and the model you can buy costs twice what Opus 4.8 does. Run your real workloads against both tiers in the free window, route by task, and treat the asterisks as the line between marketing and what you will ship.
Predictive Tech Labs helps engineering and product teams evaluate frontier models against their specific workloads and design agentic delivery pipelines that route work to the right model. Get in touch to talk through a Fable 5 vs Opus 4.8 plan for your stack.
References & Further Reading
- Anthropic (2026). Claude Fable 5 and Claude Mythos 5. anthropic.com/news/claude-fable-5-mythos-5
- Anthropic (2026). Introducing Claude Fable 5 and Claude Mythos 5 — Claude API Docs. platform.claude.com/docs
- CNBC (2026). Anthropic releases Mythos-like AI model to the public, Claude Fable 5. cnbc.com
Deciding Between Fable 5 and Opus 4.8?
We help engineering and product teams benchmark frontier models against their real workloads, model token-cost trade-offs across tiers, and design agentic pipelines that route each task to the right model.