Fable is not meaningfully different than Opus 4.8 for cybersecurity or for anything. It was just a sharper, ~10% better model. It makes a significant difference but isn’t going to suddenly hack a nuclear power plant. And now we have the government involved because of Dario’s doomer marketing and refusal to work with the government. Idiotic.
Microsoft invested $13 billion in OpenAI.
Invested $8 billion in Anthropic.
Banned its own employees from using Claude Fable 5.
Built Copilot Cowork on top of Claude.
Now wants to replace Claude with DeepSeek.
A Chinese AI company.
To cut costs.
Because their own employees were running so many tasks the bills got “bonkers.”
They called it tokenmaxxing.
They shut down the internal leaderboard that caused it.
The tokenmaxxing continued anyway.
Microsoft spent years warning about Chinese AI data risks.
Is now self-hosting DeepSeek on Azure.
Wrapped in compliance controls.
And calling it enterprise-safe.
The US government just shut down Anthropic’s models over national security.
Microsoft’s response is to switch to the Chinese model.
Sam Altman has not responded.
Dario Amodei has not responded.
The Jevons paradox means cheaper AI just means more AI.
Microsoft already knows this.
They tokenmaxxed themselves into this corner.
Seeing so many people saying Fable wasn’t that much better than other models.
Sure, if you’re just doing normal frontend/backend work that Opus/GPT 5.5 can already do, it’s going to feel basically the same.
The magic comes when you’re asking it to do things other models can’t.