#AnthropicReleasesFable5Model

Claude Fable 5 is Anthropic's most capable publicly available model and the first broadly accessible model in its new Mythos class. It is positioned for software engineering, knowledge work, long-horizon tasks, and visual reasoning.

Anthropic reported an 80.3% score on SWE-Bench Pro, a benchmark that evaluates real-world software engineering tasks.

The model includes stronger safety controls than the restricted Claude Mythos 5 version. In some high-risk domains (such as advanced biology or cybersecurity), requests may be blocked or routed to a less capable model.

API pricing is $10 per million input tokens and $50 per million output tokens, approximately double the pricing of Anthropic's earlier Opus-tier models.

Why the launch matters

The release highlights a growing trend among frontier AI labs: pushing capability higher while simultaneously adding more restrictive safeguards. Anthropic argues that Fable 5 can handle longer, more autonomous workstreams than previous public models, but it also introduced mechanisms that intentionally limit certain sensitive capabilities.

The trade-off

Stronger coding and reasoning performance Higher token costs

Better long-running task execution More aggressive safety restrictions

Public availability Some queries may be redirected or refused

Strong benchmark results Benchmark leadership doesn't always translate directly to every real-world workflow

Overall, the launch positions Claude Fable 5 as a premium, high-capability model aimed at organizations that value performance on complex engineering and analytical tasks and are willing to pay significantly more for it.