Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling

Benchmark Model Rank Results
mathematical-reasoning-on-amo-benchFalcon-H1R (7B)#4Accuracy: 36.3
mathematical-reasoning-on-aime-2025Falcon-H1R (7B) with DeepConf@512#5Accuracy: 96.7
mathematical-reasoning-on-hmmt-feb-2025Falcon-H1R (7B)#10Accuracy: 64.9
mathematical-reasoning-on-math-500Falcon-H1R (7B)#11Accuracy: 97.4
mathematical-reasoning-on-aime24Falcon-H1R (7B)#17Acc: 88.1
code-generation-on-livecodebench-v6Falcon-H1R (7B)#35Pass@1: 68.6
instruction-following-on-ifbenchFalcon-H1R (7B)#35Accuracy: 53.4
question-answering-on-humanity-s-last-exam-hleFalcon-H1R (7B)#38Accuracy: 11.1