Claude Fable 5.1 launched this week as Anthropic’s most advanced model for coding and knowledge work. Alongside it, Anthropic released Claude Mythos 5.1, a second version built on the exact same underlying model but with different safety guardrails. If you’ve seen both names floating around, this guide breaks down how they differ using Anthropic’s own published benchmarks.
Claude Fable 5.1 and Claude Mythos 5.1 are the same model with different safeguard levels. Fable 5.1 is generally available to everyone today. Mythos 5.1 is restricted to vetted cybersecurity and life-sciences professionals through Anthropic’s trusted access programs.
What Is Claude Fable 5.1, Really?
Think of Anthropic’s Claude lineup like clothing sizes: Haiku is small, Sonnet is medium, and Opus is large. Above Opus, meanwhile, sits a tier Anthropic calls Mythos-class, and Fable is the generally-available version of that top tier. Because Fable 5.1 and Mythos 5.1 share the same underlying weights, the real difference between them comes down to which safety filters are active, not raw capability.
Fable 5.1 vs. Fable 5 vs. Opus 5 vs. GPT-5.6 Sol
To back up its claims, Anthropic published head-to-head benchmark scores comparing Fable 5.1 against its predecessor and against competing models. Here’s how they stack up on a few key tests:
| Benchmark | Fable 5.1 | Fable 5 | Opus 5 | GPT-5.6 Sol |
|---|---|---|---|---|
| Agentic coding (Terminal-Bench 4.0) | 55.8% | 42.0% | 52.3% | 37.3% |
| Agentic coding (CursorBench 3.2.0) | 73.4% | 70.5% | 70.0% | 67.2% |
| Multidisciplinary reasoning (HLE, no tools) | 60.9% | 57.8% | 56.6% | — |
| Business workflows (AutomationBench) | 31.4% | 17.1% | 26.9% | 19.6% |
These figures come directly from Anthropic’s own published announcement. As with any vendor-reported benchmark, it’s worth testing a model against your own real workload rather than trusting a leaderboard alone.
What Changed for Existing Users
Beyond raw performance, Anthropic focused this release on three practical complaints from customers: price, data retention, and false-positive safeguard triggers. First, cache reads now cost 75% less — just $0.25 per million tokens. As a result, typical workloads cost around 25% less overall. Highly agentic, context-heavy work, meanwhile, can cost up to 45% less. Base pricing itself stays the same as Fable 5: $10 per million input tokens and $50 per million output tokens.
Second, Anthropic reduced how often its safety filters flag harmless requests. In cybersecurity specifically, the newest safeguards trigger 60% fewer false positives than before. This is partly because Fable 5.1 can now help identify software vulnerabilities — though it still won’t help write exploits for them. Meanwhile in biology, benign questions about elementary topics now trigger safeguards 85% less often than they did at Fable 5’s original launch.
How to Start Using Claude Fable 5.1
Step 1: Chat Directly
Go to claude.ai and start a new conversation — it’s already live there for eligible plans, with no extra setup needed.
Step 2: Build With the API
Head to platform.claude.com, create or open a developer account, and reference the model as claude-fable-5-1 in your API calls.
Step 3: Use It on Your Existing Cloud
Alternatively, Fable 5.1 is also available today through Amazon Web Services, Google Cloud, and Microsoft Azure, so teams already committed to one of those platforms don’t need a separate Anthropic billing relationship.
Step 4: Apply for Mythos 5.1 Access
If your work genuinely requires the loosened cybersecurity or life-sciences safeguards, apply through the Cyber Verification Program or the Life Sciences Verification Program rather than trying to work around Fable 5.1’s restrictions.
By default, Fable 5.1 runs at High effort inside Claude Code, but Medium effort inside Claude Cowork and on claude.ai. So if a task feels slower than expected, check which effort level you’re running before assuming something’s wrong.
The Fable vs. Mythos Decision at a Glance
- Everyday coding, writing, and research? Use Claude Fable 5.1 — it’s generally available.
- Defensive cybersecurity work needing fewer restrictions? Apply to the Cyber Verification Program.
- Life sciences R&D needing advanced biology capabilities? Apply to the Life Sciences Verification Program.
A Note on Transparency and Watermarking
Because Anthropic signed the EU AI Act’s Code of Practice on Transparency of AI-Generated Content, Fable 5.1’s outputs carry an invisible watermark, as required for models released after August 2, 2026. However, this watermark doesn’t affect output quality and contains no information about you or your conversations — it exists purely so eligible regulators, researchers, and fact-checkers can verify whether Claude likely generated a given piece of text.
Frequently Asked Questions
Is Claude Fable 5.1 free to use?
Chatting at claude.ai requires an eligible paid plan for Fable-tier access; API usage is billed per token at $10 per million input tokens and $50 per million output tokens.
Can regular developers access Mythos 5.1?
Not directly. Mythos 5.1 is currently limited to vetted US organizations through Anthropic’s trusted access programs, though Anthropic says it’s coordinating with the US government to expand access over time.
Does Fable 5.1 actually outperform Fable 5?
Yes, according to Anthropic’s own benchmarks across coding, reasoning, and business-workflow tests — and notably, it often matches or beats Fable 5’s results even at lower effort settings, which also lowers cost.