Anthropic unveils Claude Fable 5 AI model with enhanced coding, vision

Image credit: Search Engine Journal

Anthropic on Monday unveiled Claude Fable 5, its new flagship artificial intelligence model, demonstrating advanced capabilities across complex coding, knowledge work, and vision tasks globally.

The model is designed for longer, more intricate assignments, potentially transforming web development and research workflows with its enhanced performance and new safety protocols.

Claude Fable 5 showed significant improvements in coding, capable of autonomous work and complex tasks, according to Anthropic. The company reported that the model could manage a 50-million-line Ruby codebase migration in a single day.

The AI achieved the highest score on Cognition’s FrontierCode evaluation for difficult coding tasks. It also demonstrated the ability to rebuild web application source code from screenshots, a feature noted by Jamie Marsland.

In knowledge work, Claude Fable 5 excelled, scoring highest on Hebbia’s Finance Benchmark. Anthropic stated that the model showed strong performance in analytical tasks suitable for senior-level reasoning.

The new model represents Anthropic’s strongest vision model to date, offering improved long-context performance by maintaining focus across millions of tokens.

Anthropic has implemented new safety limits for Claude Fable 5. Certain sensitive requests, including those related to cybersecurity, biology/chemistry, and model extraction, are routed to Claude Opus 4.8 to prevent misuse.

The model is available on paid plans from June 10 to June 22. After this period, access will transition to usage credits, with Anthropic aiming to restore it as a standard subscription feature in the future.

For developers accessing the model via API, pricing is set at $10 per million input tokens and $50 per million output tokens, the company said.

Source: Search Engine Journal