How I AIClaude Fable 5 (Mythos) - is the world’s best coding model as good as they say?
At a glance
WHAT IT’S REALLY ABOUT
Testing Anthropic’s Claude Fable 5: benchmark beast, costly, uneven usability
- Fable 5 is positioned as the first general-availability Mythos-class model with standout benchmark results (notably SWBench Pro) but with high token consumption and premium pricing.
- Anthropic’s safety approach adds domain classifiers (cybersecurity, bio, chemistry, distillation) and a “fallback” behavior that silently routes flagged requests to Opus 4.8 rather than hard-refusing.
- In hands-on use, Fable 5 shows strong vision/document-layout ability and thorough “seasoned engineer” reasoning, but its verbose, detail-heavy prose can be difficult to parse for specs and strategy work.
- The model appears conservative when asked to execute toward an MVP and can under-deliver on ambition unless prompted carefully, potentially influenced by safety tuning and “minimal” interpretations.
- Multi-agent and long-running workflows look promising in concept, but the reviewer encountered stalls/bugs in orchestration (likely tool/harness issues), highlighting reliability as the gating factor for days-long tasks.
IDEAS WORTH REMEMBERING
5 ideasFable 5’s cost profile changes when it’s worth using.
At $10/input token and $50/output token and reportedly ~2x token usage, it’s best reserved for tasks where extra thoroughness pays off; otherwise Sonnet/Opus may be more cost-effective.
Benchmark wins don’t guarantee better day-to-day product work.
Despite “crushing benchmarks,” the reviewer found spec/prose output overly dense and hard to interpret, which can slow teams that need clarity more than exhaustive detail.
The new safety “fallback” is a pragmatic middle ground.
Instead of blocking, classified requests fall back to Opus 4.8, preserving continuity for developers while still limiting high-risk capabilities; most sessions reportedly never trigger it.
Use Fable 5 as an ‘engineer’s engineer’—but don’t expect product sense by default.
It investigates exhaustively and verifies heavily, which helps correctness, but it can over-optimize details at the expense of shipping, requiring stronger product framing in prompts.
Vision/document formatting is a standout strength.
In simple but telling tests (e.g., handwriting worksheet layout), Fable 5 produced clearer spacing and readability than Opus 4.8, suggesting real gains for PDF/layout-centric tasks.
WORDS WORTH SAVING
5 quotesIt's here, the model, the myth, the legend. Mythos from Anthropic has finally dropped. Well, baby Mythos. We're calling it Fable 5, and this new model is crushing benchmarks, but the question is, can it crush my backlog?
— Claire Vo
It's $10 per input token and $50 per output token. It's gonna be a new tier above Opus, and so if you're gonna use this model, you're gonna pay, pay the price.
— Claire Vo
They explicitly say that Fable works like a seasoned engineer. Unfortunately, if you have worked with a seasoned engineer, you know there's good to this and you know there's bad to this.
— Claire Vo
This is a real struggle with these more intelligent frontier models, is they're, like, too smart, and so it's just very, very hard to parse what they're saying.
— Claire Vo
I cannot wait to hear what you build, what you overbuild, and what you make ugly with this new model.
— Claire Vo
High quality AI-generated summary created from speaker-labeled transcript.
