Claude Fable 5.1 tops Senior SWE-Bench, matching Fable 5 at ~50% cost via Medium effort tier
ajratner · x · 2026-09-03
Claude Fable 5.1 now tops the Senior SWE-Bench leaderboard, edging out Fable 5 on the pass^3 tie-breaker (more consistently shipping quality code). Notably, its best effort tier is Medium, and combined with cheaper cache reads it matches Fable 5's performance at 50% of the cost — a lesson that medium effort is often all you need.
More from Models
- Rumor: Meta 'open weights release' incoming, Zuckerberg to reclaim open-source crown? — airesearch12 · 2026-09-03
- Could the 'Astra is a looped model' story be a deliberate red herring? — max_paperclips · 2026-09-03
- GLM-5.3-Flash priced at 0.06x: Factory confirms all-day droid usage without rate limits — matanSF · 2026-09-03
- Claude user burns $100 of usage credits in 15 minutes after hitting limits twice in a day — BLUECOW009 · 2026-09-03
- Skeptic dissects Astra's rumored recurrent architecture: likely just looping each layer twice — mike64_t · 2026-09-03
- Rumor: OpenAI's Astra reportedly launching at 1pm PST tomorrow amid cryptic teases — dolo937 · 2026-09-03