Skip to content
forward pass

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. The organism has the same path lit. Direct influences are marked.

00 · One neuron · 1

  1. 1943A logical calculus of nervous activity

I · Foundations · 11

  1. 1948A mathematical theory of communication
  2. 1949Cells that fire together wire together
  3. 1950Programming a computer for playing chess
  4. 1950Computing machinery and intelligence
  5. 1958The perceptron learns
  6. 1959Samuel's checkers program coins 'machine learning'
  7. 1960ADALINE and the least-mean-squares rule
  8. 1965Moore's law
  9. 1966ELIZA
  10. 1969Perceptrons
  11. 1970Reverse-mode automatic differentiation

W1 · The first winter · 2

  1. 1974Werbos applies backpropagation to neural networks
  2. 1980The Neocognitron

II · Connection · 3

  1. 1982The Hopfield network
  2. 1985The Boltzmann machine
  3. 1986Backpropagation

W2 · The second winter · 6

  1. 1988Temporal-difference learning
  2. 1989Q-learning
  3. 1989LeNet reads handwritten postcodes
  4. 1990Finding structure in time
  5. 1991The vanishing gradient problem
  6. 1992TD-Gammon reaches world-class backgammon

III · Statistics and data · 10

  1. 1997Long short-term memory
  2. 1998MNIST and LeNet-5
  3. 1999The first GPU
  4. 2003A neural probabilistic language model
  5. 2006Deep belief networks and the word 'deep'
  6. 2007CUDA
  7. 2009Deep learning moves to GPUs
  8. 2009ImageNet
  9. 2010Rectified linear units
  10. 2010DeepMind is founded

IV · Deep learning · 11

  1. 2012Google Brain's network discovers cats
  2. 2012Dropout
  3. 2012AlexNet wins ImageNet
  4. 2013Deep Q-networks play Atari
  5. 2014Google buys DeepMind
  6. 2014Superintelligence
  7. 2014Attention
  8. 2014Sequence to sequence learning
  9. 2015Batch normalisation
  10. 2015Residual networks
  11. 2015OpenAI is founded

V · Transformers · 10

  1. 2017The Asilomar AI Principles
  2. 2017Attention is all you need
  3. 2017Deep reinforcement learning from human preferences
  4. 2018GPT: generative pre-training
  5. 2019GPT-2 and the model too dangerous to release
  6. 2020Scaling laws for neural language models
  7. 2020GPT-3
  8. 2020Learning to summarise from human feedback
  9. 2021Anthropic is founded
  10. 2021GitHub Copilot writes code

VI · Everyone · 10

  1. 2022InstructGPT
  2. 2022ChatGPT
  3. 2023Claude
  4. 2023GPT-4
  5. 2023'Pause Giant AI Experiments'
  6. 2023The US executive order on AI
  7. 2023The Bletchley Declaration
  8. 2024Claude 3 catches GPT-4
  9. 2024The EU AI Act enters into force
  10. 2024Claude learns to use a computer

VII · Agents · 3

  1. 2025Claude 4 and Claude Code
  2. 2025America's AI Action Plan
  3. 2026Claude Fable 5 and the Mythos classdirect

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 corpus has built on it yet.

sources · 2

Trace the lineage of this event on the timeline →Back to the ledger