ScreamingFace launches leaderboards showing multi-model systems can beat the best single model inside them
iamtrask · x · 2026-11-05
ScreamingFace has launched what it calls the first verified map of the quality–cost frontier for multi-model systems (ensembles, routers, cascades), arguing that a handful of AI models combined can beat the best single model inside them.
- The leaderboard at leaderboard.screamingface.ai covers 10+ benchmarks including ARC-AGI-2, GDPval Text Subset, HealthBench Professional, MMLU-Pro, FrontierScience and InfiniteBench Math.Find, each flagged "Best Reproducible".
- Every score, recipe and cost is published to make the cost-quality pareto frontier reproducible.
Related event: OpenMined Launches ScreamingFace, an Open Fusion Model Ecosystem(2 posts)→
More from Models
- Researcher claims Claude sandbagging hampers safety and interp research — repligate · 2026-11-05
- Sol 6.1 hits 24.5 tps vs promised 750 tps at $60/1M output tokens — NandaVegg · 2026-11-05
- Unverified claim: an upcoming "Luna-level" model with 0.3B active params at $0.01/Mtok — teortaxesTex · 2026-11-05
- Toloka tests GPT-6 Astra for robot control: 90% success vs 30% for classical stack — MParakhin · 2026-11-05
- Users complain GPT-6.2 feels nerfed in unexplained quality drop — CtrlAltDwayne · 2026-11-05
- OpenAI's odd GPT-6.2 naming sparks speculation it is sandbagging version numbers — scaling01 · 2026-11-05