Coding with Claude Fable 5: Benchmarks, the Stripe Migration, and Real Workflows
What Claude Fable 5 changes for software engineering: an 80.3% SWE-Bench Pro score, a 50-million-line migration done in a day, and new agentic workflows.
TL;DR
Claude Fable 5 is the strongest coding model released to date: 80.3 percent on SWE-Bench Pro versus 69.2 percent for Opus 4.8, 58.6 percent for GPT-5.5, and 54.2 percent for Gemini 3.1 Pro, plus the top frontier-model score on Cognition's FrontierCode eval for code quality and efficiency. In an early test, Stripe completed a 50-million-line Ruby codebase migration in one day that would have taken a team over two months.
The Benchmark Picture
SWE-Bench Pro measures whether a model can resolve real software engineering tasks end to end. Fable 5's 80.3 percent is roughly 11 points ahead of the next best model. FrontierCode adds a second dimension - not just whether the task got done, but how good and how efficient the resulting code is - and Fable 5 leads frontier models there too.
The qualitative reports match. Cursor CEO Michael Truell: "Claude Fable 5 is the state of the art model on CursorBench. It's opened up a class of long-horizon problems that were out of reach." Lyzr CTO Fabian Hedin: "Apps that took a hundred prompts a year ago, it now one-shots."
The Stripe Migration
The most striking early result is Stripe's: a migration across a 50-million-line Ruby codebase completed in one day, against an internal estimate of more than two months for an engineering team. Anthropic described it as having "compressed months of engineering into days." Two Fable 5 traits make this possible:
- Long-context endurance - the model maintains focus across millions of tokens, so it does not lose the thread halfway through a codebase-scale task
- Fewer correction turns - as Equinox CTO Luke Anderson noted, it "delivers more capable engineering in fewer turns than prior models"
What Changes in Daily Workflows
The practical shift is the size of the unit of work you can hand off. With prior models, the safe delegation unit was a function, a file, or a small PR. With Fable 5, early users are delegating multi-day refactors, dependency upgrades, and migrations as single tasks with a clear spec up front. GitHub CPO Mario Rodriguez framed the trajectory: "What excites us most is the direction it points: a future where developers can hand increasingly ambitious work to agents."
Concrete patterns worth trying during the June 9-22 free window on paid Claude plans:
- Give the full task specification in one well-written brief, then let the model run long rather than steering turn by turn
- Use file-based memory for multi-session projects - in a Slay the Spire test with file-based memory, Fable 5 performed 3x better than Opus 4.8
- Point it at real migration backlogs: framework upgrades, API deprecations, and type-system adoption are exactly the long-horizon shape it excels at
Availability for Developers
Fable 5 is fully available on the Claude API at 10 dollars per million input tokens and 50 per million output, and generally available on Amazon Bedrock and GitHub Copilot - so it slots into existing AWS and GitHub-based development stacks immediately.