Anthropic ships Claude Fable and Mythos 5.2 with faster inference

The release notably omits any mention of how these models handle the financial integrations Anthropic shipped three days ago (bank account access via Plaid). If Fable and Mythos 5.2 don't materially improve financial reasoning or transaction safety, the timing suggests Anthropic is decoupling capability releases from the guardrails required for high-stakes use cases.

This lands amid a much larger acceleration narrative. METR's November 7 report found Anthropic's development velocity has doubled since July, with AI-assisted research driving roughly 80 percent of gains. Today's incremental update fits that pattern, but it also raises a question the summary glosses over: are these quarterly bumps the output of the automated pipeline METR documented, or are they still human-driven? The bank account feature from November 5-6 tested whether Claude could handle sensitive data reliably. These new models should be the proof point, yet Anthropic hasn't claimed improved financial reasoning specifically.

If Anthropic publishes financial-domain benchmarks (accuracy on transaction categorization, spending pattern inference, or fraud detection) within 30 days, that confirms Fable and Mythos 5.2 were built to support the banking integration. If no such benchmarks surface by mid-December, it signals the models were released on a separate cadence and the bank feature is still running on older capability foundations, which would undermine claims about integrated product velocity.