> ## Content Index
> Fetch the complete content index at: https://www.metatalks.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Anthropic releases Claude Fable 5.1 and Mythos 5.1, one model in two editions
- URL: https://www.metatalks.ai/anthropic-fable-5-1-mythos-5-1-one-model-two-editions/
- Published: 2026-09-01T22:52:41.000Z
- Updated: 2026-09-01T22:54:12.000Z
- Author: Al
- Tags: News, Frontier Models, AI safety, #newswire

**Fable 5.1 beats the cheaper Opus 5 on every published benchmark, cuts the price of cached input by three quarters, and may now be pointed at finding software vulnerabilities — though still not at writing exploits.**

Anthropic released [Claude Fable 5.1 and Claude Mythos 5.1](https://www-cdn.anthropic.com/0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20&%20Claude%20Mythos%205.1%20System%20Card.pdf?ref=metatalks.ai) on September 1 — the same model, by the company’s own account, with different levels of safeguards layered on top. Fable 5.1 went to everyone that day, through cloud providers or by calling the API. Mythos 5.1 goes only to vetted practitioners in cybersecurity and the life sciences, so far only at US organizations, through two verification programs that take over from Project Glasswing.

Opus 5, the cheaper model that arrived in July, had been beating the flagship on most of the benchmarks Anthropic publishes. On agentic coding, measured with Terminal-Bench 4.0, [Anthropic’s announcement](https://www.anthropic.com/claude-fable-and-mythos-5-1?ref=metatalks.ai) now puts Fable 5.1 at 55.8% and Mythos 5.1 at 60.9%, against 52.3% for Opus 5 and 42.0% for Fable 5.

The list price has not moved: $10 per million input tokens and $50 per million output, twice what Opus 5 charges at $5 and $25\. What changed is cached input, which now costs $0.25 per million — enough, on Anthropic’s estimate, to make a typical workload around 25% cheaper to run than Fable 5, and heavily agentic work up to 45% cheaper. Set the model to low or medium reasoning effort, the company says, and it matches or beats what Fable 5 used to deliver.

Anthropic has also loosened the safeguards where it says they were catching the wrong work. On its numbers, people working in Claude Code should see roughly 60% fewer safeguard interruptions in a session than the earlier protections on Fable 5 produced. Penetration testing, exploit writing and vulnerability hunting in binaries still go to its Opus models.

Enterprise Frontier Safeguards, Anthropic’s answer on data retention, is not ready. It keeps the customer’s data on cloud infrastructure the customer runs rather than the vendor’s, and leaves any human review to the customer by default; it arrives in stages from this autumn. Until then, customers who qualify can run Fable 5.1 under an arrangement in which no data is retained.