Updated October 2026. Claude Fable 5.1 is Anthropic’s most capable generally available model, released on September 1, 2026 alongside Claude Mythos 5.1. This Claude Fable 5.1 review covers what actually changed from Fable 5, the benchmark numbers Anthropic published, what it really costs to run through the API, and — just as important — when you should pick a cheaper Claude model instead.
Table of Contents
Claude Fable 5.1 at a Glance
| Spec | Claude Fable 5.1 |
|---|---|
| Release date | September 1, 2026 |
| API model ID | claude-fable-5-1 |
| Input price | $10 per million tokens |
| Output price | $50 per million tokens |
| Cache reads | $0.25 per million tokens (75% cheaper than Fable 5) |
| Context window | 1M tokens (per Artificial Analysis) |
| Effort levels | low, medium, high, xhigh, max |
| Availability | Claude apps, Claude API, AWS, Google Cloud, Microsoft Azure |
What Is Claude Fable 5.1?
Fable 5.1 and Mythos 5.1 are the same underlying model shipped with two different safeguard levels. Fable 5.1 is the generally available version. Mythos 5.1 has more permissive safeguards for cybersecurity and life-sciences work and is only offered to vetted organizations through Anthropic’s trusted-access programs (currently US-only). If you want the background on how the Fable and Mythos lines started, see our explainer on Claude Fable 5 and Mythos 5.
Anthropic positions Fable 5.1 for coding, knowledge work, and long-running multi-step tasks: orchestrating research, incident investigation, code review, finance work, document redlining, and building presentations.
What Changed From Fable 5
- Cheaper in practice, same list price. Input and output prices did not move, but cache reads dropped 75% to $0.25 per million tokens. Anthropic estimates roughly 25% lower costs on typical workloads and up to about 45% on highly agentic ones.
- Better scores across the board on Anthropic’s published benchmarks (table below).
- Fewer false refusals. About 60% fewer cyber-safeguard interventions per Claude Code session, and biology safeguards fire 85% less often on benign elementary biology and medical questions.
- Security work. Fable 5.1 can identify software vulnerabilities, but not develop exploits. Dual-use tasks like penetration testing and exploit generation are redirected to Opus models.
- Effort control at lower cost. Anthropic says Fable 5.1 at low or medium effort matches or beats Fable 5 at much lower cost.
Claude Fable 5.1 Benchmarks
For this Claude Fable 5.1 review, these are the results Anthropic reported at launch, measured with production safeguards switched on. Treat vendor benchmarks as a starting point, not a verdict.
| Benchmark | Fable 5.1 | Fable 5 | Opus 5 | GPT-5.6 Sol |
|---|---|---|---|---|
| Terminal-Bench-Science 0.1 | 52.6% | 24.7% | 29.0% | 22.4% |
| OSWorld 2.0 (partial) | 77.9% | 72.9% | 75.4% | — |
| Humanity’s Last Exam (no tools) | 60.9% | 57.8% | 56.6% | — |
| Humanity’s Last Exam (with tools) | 65.0% | 63.8% | 63.6% | — |
| GDPval-AA v2 (Elo) | 1853 | 1723 | 1824 | 1711 |
| AutomationBench | 31.4% | 17.1% | 26.9% | 19.6% |
| CursorBench 3.2.0 | 73.4% | 70.5% | 70.0% | 67.2% |
On the independent Artificial Analysis Intelligence Index, Fable 5.1 scores 53 at max and xhigh effort, 51 at high, 49 at medium and 47 at low. Output speed sits roughly between 56 and 68 tokens per second depending on effort.
Real API Pricing: What It Costs to Run
At $10 input and $50 output per million tokens, Fable 5.1 is the most expensive model in Anthropic’s lineup. Three levers bring the real bill down:
- Prompt caching. Cache reads cost $0.25 per million tokens. If your agent re-sends the same system prompt, tool definitions or documents, caching is the single biggest saving.
- Effort level. Artificial Analysis measured about $2.37 per index task at low effort versus about $7.63 at the default setting. Start low and raise effort only for the tasks that need it.
- Batch processing for anything that does not need a real-time answer.
If you are new to the API, our Claude API tutorial walks through the setup, and you can check current rates on Anthropic’s Fable page.
Effort Levels and Defaults
Fable 5.1 supports five effort levels: low, medium, high, xhigh and max. Claude Code defaults to high; Claude.ai and Claude Cowork default to medium. Higher effort means more reasoning tokens — and more output billed at $50 per million — so it is worth setting effort per task instead of leaving everything on max.
Limitations to Know Before You Switch
- Price. It is the priciest Claude model, so routing every request to it rarely makes sense.
- Safeguards still apply. Offensive security tasks are redirected, and Mythos 5.1 is the only route to the more permissive version.
- Known issues. Anthropic lists occasional bypassing of approvals and auto-mode classifiers, plus limited audit coverage for very long-context and multi-agent setups.
- Anti-distillation change. New API accounts can no longer manually edit prior context while preserving the transcript of Claude’s earlier thinking.
Fable 5.1 or Opus 5.5?
Three weeks after Fable 5.1, Anthropic released Claude Opus 5.5 (September 22, 2026) at $4 input and $20 output per million tokens. Anthropic says Opus 5.5 “performs at the level of Claude Fable 5.1 on most work”, and it leads Fable 5.1 on Anthropic’s agentic coding, computer use and knowledge-work benchmarks. For most teams, that makes Opus 5.5 the default and Fable 5.1 the specialist. We compare them head to head in Claude Fable 5.1 vs Opus 5.5.
Claude Fable 5.1 Review: Verdict
Claude Fable 5.1 is a real step up from Fable 5: better scores, fewer false refusals, and meaningfully cheaper once caching is in place. But with Opus 5.5 matching it on most work at well under half the price, Fable 5.1 is now best reserved for the hardest long-running tasks where you have measured a quality gain on your own prompts. Run both on a sample of real traffic, compare cost per successful result, and route accordingly.
Sources: Anthropic — Introducing Claude Fable 5.1 and Mythos 5.1, Anthropic — Claude Opus 5.5, Artificial Analysis.