Claude Fable 5.1 and Claude Mythos 5.1: Anthropic's New Standard for Coding and Scientific Research
- 3 days ago
- 4 min read
Anthropic has announced Claude Fable 5.1 and Claude Mythos 5.1, introduced as its most advanced models yet for coding and knowledge work, with research results that hint at where AI-driven science is headed.
........
Both models share the same underlying model, but with different levels of safeguards: Fable 5.1 is generally available, while Mythos 5.1 is offered only through trusted access programs.
Fable 5.1 costs about 25% less than Fable 5 for typical workloads, and up to 45% less for highly agentic tasks.
The new Enterprise Frontier Safeguards will let companies keep their data on their own cloud infrastructure, with privacy equivalent to zero data retention.
On agentic coding and multidisciplinary reasoning benchmarks, Fable 5.1 outperforms both Fable 5 and Opus 5.
Mythos 5.1 delivered record results in molecular design and in optimizing computational models for biology research.
··········
WHAT CHANGED WITH THE 5.1 RELEASE.
Fable 5.1 and Mythos 5.1 are the same underlying model, differing only in the level of safeguards applied.
Fable 5.1 is the generally available version, built for coding, knowledge work, and long-running research tasks.
Mythos 5.1 is instead reserved for trusted access programs, with safeguards specifically designed around cybersecurity and life sciences work.
Alongside the capability upgrade, Anthropic also addressed three recurring customer complaints: price, data retention, and overly aggressive safeguards.
··········
PRICING AND USAGE COSTS.
The price reduction mainly comes from cache reads, which now cost 75% less.
For a typical workload, overall costs drop by around 25% compared with Fable 5.
For highly agentic, tool- and context-heavy tasks, savings can reach up to 45%.
Base pricing is unchanged from Fable 5: $10 per million input tokens and $50 per million output tokens.
Cache reads instead drop to $0.25 per million tokens.
··········
Indexed cost comparison (Fable 5 = 100)
Workload type | Fable 5 | Fable 5.1 | Estimated savings |
|---|---|---|---|
Typical workload | 100 | 75 | ≈25% |
Highly agentic workload | 100 | 55 | ≈45% |
··········
BENCHMARK PERFORMANCE.
Fable 5.1 posts its strongest scores yet on agentic coding, scientific research, and multidisciplinary reasoning.
On the Terminal-Bench-Science agentic research benchmark, Fable 5.1 scores 52.6%, compared with 24.7% for Fable 5 and 29.0% for Opus 5.
On agentic coding (Terminal-Bench 4.0), Fable 5.1 reaches 55.8%, rising to 60.9% for the Mythos 5.1 variant.
On multidisciplinary reasoning (Humanity's Last Exam, no tools), Fable 5.1 hits 60.9%, ahead of Fable 5 (57.8%) and Opus 5 (56.6%).
··········
Key benchmark comparison
Benchmark | Fable 5.1 | Fable 5 | Opus 5 |
|---|---|---|---|
Terminal-Bench-Science 0.1 | 52.6% | 24.7% | 29.0% |
Terminal-Bench 4.0 (agentic coding) | 55.8% | 42.0% | 52.3% |
Humanity's Last Exam (no tools) | 60.9% | 57.8% | 56.6% |
CursorBench 3.2.0 | 73.4% | 70.5% | 70.0% |
AutomationBench | 31.4% | 17.1% | 26.9% |
··········
EARLY SCIENTIFIC RESEARCH RESULTS.
Anthropic tested the two models' research capabilities across several scientific fields, with results that go beyond lab demos.
Molecular design was one of the clearer wins: Mythos 5.1 produced high-affinity protein binders, with a hit rate close to 50% across 12 targets, compared with the 10–15% typical in the field.
Claude Fable 5.1 also trained a neural network to build a high-resolution elevation map covering a third of the planet Venus, based on radar data from NASA's decades-old Magellan mission.
On the computational biology side, Mythos 5.1 sped up seven open-source deep learning models by up to 2.5 times, cutting estimated GPU costs by 30–60% on large-scale genomic analyses.
··········
SAFETY AND SAFEGUARDS.
Anthropic ran the models through months of red-teaming across chemical, biological, cyber, and alignment risks before release.
On the cyber side, Mythos 5.1 shows the strongest capabilities Anthropic has released to date, while still falling within the lowest risk tier of its Frontier Compliance Framework.
The new cybersecurity safeguards generate around 60% fewer false interventions than Fable 5's, while still blocking dual-use activities such as penetration testing and exploit generation.
On biology, the safeguards remain those already applied to Mythos 5, restricting advanced biological research capabilities to vetted programs.
Alignment evaluations show improvement over Mythos 5, with the model less likely to bypass explicit constraints or justify actions through motivated reasoning.
··········
TECHNICAL SPECIFICATIONS AND API CHANGES.
Fable 5.1 and Mythos 5.1 share the same underlying specs, confirmed directly in Anthropic's developer documentation.
Both models run a 1 million token context window, which is also the default and the maximum, billed at standard per-token pricing across the whole window.
Maximum output is capped at 128,000 tokens per response.
Adaptive thinking is always on, and depth is controlled through the effort parameter, which can now be adjusted per message in beta.
··········
Fable 5.1 vs Fable 5: what changes for developers
Type of change | Detail |
|---|---|
Breaking | Forced tool_choice (any/tool) now returns an error |
Breaking | Earlier Claude models can no longer read Fable 5.1's thinking blocks |
Breaking | Editing earlier turns invalidates thinking blocks (applies to new API accounts created from August 31, 2026) |
Additive (beta) | Per-message effort adjustment |
Additive (beta) | Turn-scoped system messages |
Additive (beta) | Readable progress updates between tool calls |
Additive | Lower cache read pricing |
Additive | Content provenance (watermarking) |
··········
TIMELINE: FROM FABLE 5 TO FABLE 5.1.
Fable 5.1 picks up where Fable 5 left off, just a few months after its rocky start.
Fable 5 and Mythos 5 first launched on June 9, 2026, as Anthropic's first Mythos-class tier, positioned above Opus.
On June 12, 2026, Anthropic suspended access to both models to comply with US Department of Commerce export controls.
Those controls were lifted on June 30, 2026, and Anthropic restored access the following day, July 1, 2026.
Fable 5.1 and Mythos 5.1 then launched on September 1, 2026, just two months after access was fully restored.
··········
AVAILABILITY AND ACCESS.
Claude Fable 5.1 is available today across all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure.
Developers can start using it via the API under the identifier claude-fable-5-1.
Claude Mythos 5.1 remains reserved for vetted professionals through two programs: the Cyber Verification Program for defensive cybersecurity work, and the Life Sciences Verification Program for biological research.
Anthropic also announced Enterprise Frontier Safeguards, a system that will let companies keep their data on their own cloud infrastructure with privacy equivalent to zero data retention, rolling out gradually starting this fall 2026.
··········
·····
FOLLOW US FOR MORE.
·····
·····
DATA STUDIOS
·····