List your company
ResearchFormerly WizWand

Mathematical Reasoning on Beyond AIME (Accuracy (%))

58.8AccuracyGemini 2.5 pro
-2.35213.52429.445.276Sep 2, 2025Oct 4, 2025Nov 6, 2025Dec 9, 2025Jan 10, 2026Feb 12, 2026Mar 17, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Gemini 2.5 proEvaluation Protocol=pass@12025.0958.8
FIGR2025.1254
Qwen3-235B-A22BModel Category=Large Language Models, Model Scale=235B-A22B, Reasoning Strategy=Thinking2025.1252
GLM-4.5VModel Category=Large Vision-Language Models, Model Scale=108B2025.1247
Text-only RLMode=Text-only, Reasoning Strategy=RL2025.1246
Qwen3-VL-32B-InstructModel Category=Large Vision-Language Models, Model Scale=32B, Reasoning Strategy=Instruct2025.1243
Qwen3-32BModel Category=Large Language Models, Model Scale=32B, Reasoning Strategy=Thinking2025.1240
Qwen3-VL-8B-InstructModel Category=Large Vision-Language Models, Model Scale=8B, Reasoning Strategy=Instruct2025.1230
PACS-8BEvaluation Protocol=pass@1, Backbone=Qwen3-8B2025.0928.86
PACS-4BEvaluation Protocol=pass@1, Backbone=Qwen3-4B2025.0927.16
Qwen3-14BEvaluation Protocol=pass@12025.0918.59
Qwen3-4BEvaluation Protocol=pass@12025.0917.33
Qwen3-8BEvaluation Protocol=pass@12025.0916.28
Qwen3-32BModel Category=Large Language Models, Model Scale=32B, Reasoning Strategy=Non-Thinking2025.1216
Ouro2.6B-Thinking + RLTT2026.0216
LIMOEvaluation Protocol=pass@12025.0915.98
DyJRBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.05, M=82026.0312.7
OpenThinker3-7BEvaluation Protocol=pass@12025.0912.33
Ex-GRPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0311.6
DyJR (alpha_JS = 0.01)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.012026.0311.5
DyJR (Forward KL)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, Divergence Type=Forward KL2026.0311.3
DyJR (M = 16)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=162026.0311.3
DyJR (M = 32)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=322026.0310.9
DPH-RLBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0310.8
GRPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0310.7
DAPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0310.6
RLEPBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0310.6
DyJR (M = 64)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=642026.0310.5
DyJR (alpha_JS = 0.2)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.22026.0310.1
Sky-T1-7BEvaluation Protocol=pass@12025.098.54
Eurus-2-7B-PRIMEEvaluation Protocol=pass@12025.096.67
Ouro2.6B-Thinking + SFT2026.026
Ouro2.6B-Thinking + GRPO2026.026
Ouro2.6B-Thinking2026.025
Base ModelBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.034.3
DeepSeekR1 – 7B2026.023
Marco-o1Evaluation Protocol=pass@12025.092.77
DeepSeekR1 – 1.5B2026.022
Bagel-7B-MoTModel Category=Unified Multimodal Models, Model Scale=7B, Reasoning Strategy=MoT2025.121
Deepthought-8BEvaluation Protocol=pass@12025.090.76
Bagel-Zebra-CoTModel Category=Unified Multimodal Models, Model Scale=7B, Reasoning Strategy=CoT2025.120
DeepEyesModel Category=Tool-Augmented Vision-Language Models2025.120
Chain-of-FocusModel Category=Tool-Augmented Vision-Language Models2025.120
Qwen3 – 1.7B2026.020
Qwen3 – 4B2026.020