Claude Fable 5.0 sets a new frontier score on runescape bench. y axis is log scale
Users express gratitude for the Claude Fable 5.0 Runescape AI Benchmark record because it covers the only benchmark they care about.
Most Activity

Claude Fable adeptly navigates the multi-step supply chains needed for smithing and crafting, and makes sharp navigation of tradeoffs when optimizing. For instance, it's the first model I've seen use the bank to store resources

Comparing to Opus 4.8, it's about the same speed in tokens per second, which surprises me, I expected "mythos" to be giant and slow. Plus it tends to be about 20% shorter in its responses than opus! (Still comes out to 2x the cost per run) You can view the results here: https://maxbittker.github.io/runebench/

@maxbittker The only benchmark I care about: thanks for the update 🙏