Opus 5 Outperforms Fable
Anthropic just dropped Opus 5 — smaller than Fable 5 but it beats it on several benchmarks. Plus, safety classifiers trigger 85% less, no 30-day data retention, and a new fallback feature for smoother API calls. #AI #Anthropic #Opus5 #Fable5 #MachineLearning #AIModels #TechNews #
What the video says
A massive shift in AI power dynamics has arrived with a surprising twist. Anthropic has launched its new Opus 5 model, and despite being smaller than Fable 5, it actually outperforms it on several benchmarks.
The model arrives just 2 months after Opus 4.8, boasting a much lighter touch to its safety restrictions. Anthropic documented that safety classifiers will engage 85% less often on Opus 5 than they do on Fable 5.
Furthermore, privacy-conscious users will find that Opus 5 is completely free from the 30-day data retention policy that covers Fable and Mythos. To make remaining safety interventions less disruptive, Anthropic is rolling out a beta feature called automatic fallbacks.
This new tool automatically routes blocked prompts to a less powerful model, ensuring API users receive a functional response instead of an error.
