Claude Opus 4.8
vs Fable 5.

Which model builds
a better brain?

When you build a knowledge graph from content, the model doing the reading shapes everything that comes out. So I ran a head-to-head: the same 134 YouTube videos, processed by two Anthropic models, Claude Opus 4.8 and Fable 5, into a separate knowledge graph for each. Same input, two brains. Which one is more valuable to actually think with?

Same input,
two graphs.

I built one graph with each model from the identical 134-video set, then asked both the same questions and compared the answers, looking at how richly each graph connected ideas, how accurately it grounded its responses, and which one was genuinely worth the effort to build.

flowchart TD
  A["134 YouTube videos (same set)"] --> B["Build graph with Claude Opus 4.8"]
  A --> C["Build graph with Fable 5"]
  B --> D["Ask both graphs the same questions"]
  C --> D
  D --> E["Compare value: connections, grounding, usefulness"]

Two brains,
side by side.

The clearest difference was speed and autonomy. Opus 4.8 worked through all 134 videos on its own in about two and a half hours. Fable 5 took roughly eight hours for the identical job. Both produced a usable graph, and Claude was reliable at generating the nodes and edges Neo4j needs, but the cost of getting there was very different.

The full breakdown, both graphs and how their answers differ, lives on Brain Graph. See the comparison on itsbraingraph.ai →