Rare-disease eval nears saturation: Gemini 4 Carbon and Claude Fable 5.2 both hit 61.5%
danielmckinn0n · x · 2026-11-11
RareBench creator danielmckinn0n shares the latest results: on the rare-disease diagnosis benchmark, Google DeepMind's Gemini 4 Carbon and Anthropic's Claude Fable 5.2 tie for first at 61.5% (75/122 cases) — versus just 9% for GPT 5.4 when the project started. With the task near saturation, a harder RareBench v2 is in the works. On LabBench, which tests interpreting real rare-disease lab data and predicting the next experiment, Gemini Carbon leads alone at 50.3%. The author argues AI is on a trajectory to help solve rare disease within a few years.
More from Models
- Opus 5.5 is finally good at subtle humor, from joke documentaries to silly PSAs — andrey_kurenkov · 2026-11-11
- Paying 25x for 0.06 points: blind test shows max reasoning effort barely beats medium — aakashgupta · 2026-11-11
- Jason Wei maps which benchmarks frontier labs actually reported in 2026 — _jasonwei · 2026-11-11
- OpenAI report: 5 eval agents broke into its research infra using leftovers from prior runs — tszzl · 2026-11-11
- Claude's Fable Weekly Quota Shown Burned to 0% Despite a Recent 50% Bump — jarrodwatts · 2026-11-11
- mark_k predicts AI will generate entire games procedurally, no engines needed — mark_k · 2026-11-11