Claude Mythos 5

CNBC reports that Anthropic unveiled Claude Mythos in April 2026 to a select group of companies under the Project Glasswing cybersecurity initiative, limiting the rollout because of the model's advanced cybersecurity capabilities and concerns about its potential misuse. The company indicated at the time that it did not Two months after that limited debut, Anthropic announced Claude Fable 5 — a Mythos-class model — to enterprise customers and paid subscribers, framing the broader release as possible thanks to new safeguards that block responses in specific high-risk domains such as cybersecurity and biology. Anthropic's head of produc

Anthropic released Claude Mythos 5 on June 9, 2026 as the restricted counterpart to Claude Fable 5, splitting one frontier model into two distinct products. Mythos 5 shares the same underlying base model as Fable 5 but ships with some safety classifiers lifted, and access is gated to vetted partners working on cyberdef Technically, Mythos 5 carries the same pricing as Fable 5 at $10 per million input tokens and $50 per million output tokens, and benefits from a 90% prompt-cache discount on input tokens. Reported performance from Anthropic's launch materials places it at 80.3% on SWE-Bench Pro (versus 69.2% on Opus 4.8) and a GDPval-A

Artificial Analysis's June 9, 2026 article confirms that Claude Fable 5 and Claude Mythos 5 share the same underlying model, with Fable 5 adding security guardrails for potentially harmful cybersecurity, biology, chemistry, and distillation-related queries, and introducing a fallback mechanism that routes flagged queri The same article notes that Fable 5 fell back to Opus 4.8 on 2% of GDPval-AA tasks, with Anthropic stating fallback occurs in fewer than 5% of sessions on average. The piece treats Mythos 5 as the restricted counterpart rather than benchmarking it directly, and does not discuss Azure Cognitive Services hosting. It situ

BenchLM's tracking page for Claude Mythos 5 records an API model ID of claude-mythos-5 and a 1M-token context window, sourced to Anthropic's model overview. The profile is marked as superseded by Claude Mythos 5.1, with no public ranking yet, and as of September 10, 2026 it exposed 20 sourced benchmark rows across agen Pricing on the same page is listed as $10 per million input tokens and $50 per million output tokens, with a cached input price of $1, a batch-plus-cache price of $0.50, and a blended rate of $30 per million, against field medians of $0.97 input and 256,000-token context. The page does not address Azure Cognitive Servi