Claude Fable 5.1 and Mythos 5.1: cheaper cache, stronger coding

On September 1, 2026, Anthropic released Claude Fable 5.1, alongside a restricted-access sibling model, Mythos 5.1. Headline pricing didn't change from Fable 5, but cache-read cost dropped 75%, and coding/research benchmarks improved significantly. This is the kind of release that looks like "just a point version" at first glance, then turns out to matter a lot more once you measure it against actual operating cost, rather than a noisy name change.

What happened

Fable 5.1 went straight to general availability, callable via the Claude API, and through Amazon Bedrock, Google Cloud, and Microsoft Foundry. Anthropic positions the model as its "most advanced model for coding, knowledge work, and long-running problem-solving," above the Opus, Sonnet, and Haiku tiers. Mythos 5.1 launched alongside it, but is only available to registered Anthropic partners working in cybersecurity or life sciences research.

The gist in a minute

• Release: 2026-09-01
• Headline pricing unchanged: $10/M input, $50/M output tokens
• Cache reads: $1 -> $0.25 per million tokens (75% cut)
• Estimated savings: roughly 25% typical, up to 45% for complex/agentic work
• Sibling model: Mythos 5.1, restricted to registered cybersecurity/life-sciences partners
• Benchmark: Terminal-Bench-Science 0.1, more than double Fable 5's score; Terminal-Bench 4.0: 55.8% vs. Fable 5's 42.0%

Why the cache price is what actually matters

Item	Fable 5	Fable 5.1	Change
Input	$10 / M tokens	$10 / M tokens	none
Output	$50 / M tokens	$50 / M tokens	none
Cache read	$1 / M tokens	$0.25 / M tokens	-75%

In a long-running agentic task, where the model re-reads the same large context (codebase, docs, prior steps) over and over, cache reads dominate the actual bill, not the raw input/output price. That's why the 75% cache cut delivers a much bigger real-world saving than the headline pricing would suggest on its own.

"Same price, just the cache got four times cheaper: like the bread at the store staying the same price while the shelving fee drops to a quarter. You still pay both in the end, just less of one." (Nasus)

Who benefits

For teams already running long, agentic Claude workflows (codebase analysis, code review, multi-step research), Fable 5.1 is essentially a free performance upgrade: prompts and integration don't change, you just swap the model ID for claude-fable-5-1. Mythos 5.1 targets a narrow audience: if your team isn't a registered cybersecurity or life-sciences partner, this model isn't available yet, and it's not worth planning around.

Source: Anthropic: Introducing Claude Fable 5.1 and Claude Mythos 5.1, TechCrunch: Anthropic's new Fable release is cheaper, less restrictive, Bloomberg: Anthropic Says New Fable 5.1 AI Model Is Cheaper, Better at Coding