Skip to content
The Shape of Intelligence

Claude Fable 5.1

Anthropic's Fable 5.1 and Mythos 5.1 arrive cheaper and with safeguards that block far fewer legitimate requests; the model is permitted to find software vulnerabilities but not to exploit them.

category
model
significance
3 of 5
people
Dario Amodei
organisations
Anthropic

what had to happen · 67 events back to 1943

Every event this one built on, transitively, in order. Direct influences are marked.

Three months after the June release and its interruption, Anthropic shipped the point release on 1 September 2026. Fable 5.1 and Mythos 5.1 are one model with two levels of safeguard: Fable generally available, Mythos through the company's trusted-access programmes for cybersecurity and the life sciences. The company said Fable 5.1 performed markedly better than Fable 5 on the same tasks for about a quarter less cost, and that its cyber classifiers now blocked sixty percent fewer legitimate requests. The line it drew was explicit: the model could be used to discover vulnerabilities in software, not to develop exploits for them.

The summer had also filled out the family. Sonnet 5 on 30 June replaced Sonnet 4.6 as the default model in Anthropic's products, and Opus 5 on 24 July brought the Opus tier close to Fable's frontier at half the price with a million tokens of context. The model that wrote most of this site is Fable 5.1.

The release is on this timeline as the current state of the frontier at the time of writing, and as an example of the shape the safety debate had taken by 2026: not whether a model could do harm, which was settled, but where exactly the classifier should draw the line.

what it led to · 0 events downstream

A leaf, for now. Nothing in the archive has built on it yet.

sources · 2

See this era in the exhibition →Back to the timeline