Claude Fable 5 Unveiled: Record-Breaking Performance and New Pricing

Anthropic announced the release of Claude Fable 5, stating that it outperforms all previously published models and leads on almost every AI benchmark, including software engineering, knowledge work, vision, and scientific research.

The model is more token‑efficient, handling millions of tokens in long‑running tasks without losing focus, and it improves its output using self‑generated notes. For safety, Anthropic routes certain queries or suspected distillation attempts to Claude Opus 4.8, with about 95% of conversations remaining at full capability.

Pricing is set at $10 per million input tokens and $50 per million output tokens—roughly half the cost of the Claude Mythos preview but double the price of Opus 4.8 and higher than GPT‑5.5 for input and output. From today until June 22, Pro, Max, Team, and seat‑based enterprise subscribers can use Fable 5 for free; after that date the model will be removed from those plans.

Benchmark tables (see images) show Fable 5 achieving “crushing” scores across dimensions. In a software‑engineering scenario, the model migrated a 50‑million‑line Ruby codebase in a single day, a task that would normally require a team of two months. In the FrontierCode evaluation, Fable 5 earned the highest score with a medium‑effort setting, demonstrating superior token‑usage efficiency.

On knowledge‑work benchmarks, Fable 5 topped the Finance Benchmark from Hebbia, excelling in document‑based reasoning, chart and table interpretation, and problem solving. For vision tasks, it set a new SOTA by extracting precise values from scientific charts and reconstructing web‑app source code from screenshots, and it succeeded in playing Pokémon FireRed with only minimal visual scaffolding.

The model also shows strong capabilities in long‑context memory, drug design, molecular biology, and genomics. When playing the card‑building game Slay the Spire, granting the model persistent file‑level memory access boosted its performance threefold over Opus 4.8 and allowed it to reach later game stages more frequently.

Anthropic also released Claude Mythos 5, built on the same base as Fable 5 but with fewer safety restrictions. Internal protein‑design experts reported a ten‑fold efficiency gain in certain workflow steps. Mythos 5 is the first Anthropic model to consistently generate novel, convincing scientific hypotheses; in blind comparisons with Opus‑level models, scientists preferred Mythos‑proposed hypotheses about 80% of the time, and some hypotheses have already been experimentally validated.

In a recent report, Anthropic warned that large‑model self‑iteration is accelerating and called for a global pause on AI development. Accordingly, they imposed a restriction that Fable 5 cannot be used to develop new large models, positioning this limitation as a way to deepen their competitive moat.