AIFable 5.1 reportedly debuts at No. 2 on RSI-Exam, behind GPT-6-astra
A September 18, 2026 post puts Fable 5.1 at 0.4813 behind GPT-6-astraβs 0.5126, and says no model had yet reached the benchmarkβs frontier-calibrated reference.
TLDR
RSI-Exam says it tests whether an AI agent can improve a working method and generalize to unseen data. It lists 88 active tasks across six domains: 35 public and 53 private. A September 18, 2026 post announces three leaderboard additions: Fable 5.1, Muse Spark and Seed-Evolving-0909. It says GPT-6-astra remains first at 0.5126, with Fable 5.1 entering second at 0.4813. According to the post, no model had reached the frontier-calibrated reference as of that release.
Combined views
6.1K
3 Sources, first seen 5h ago
Fable 5.1 reportedly debuts at No. 2 on RSI-Exam, behind GPT-6-astra
A September 18, 2026 post puts Fable 5.1 at 0.4813 behind GPT-6-astraβs 0.5126, and says no model had yet reached the benchmarkβs frontier-calibrated reference.