List your company
ResearchFormerly WizWand

Math Reasoning on AIME 2024 (Accuracy)

95AccuracyGPT-5 high
8.1630.70553.2575.795Jul 2, 2025Aug 22, 2025Oct 13, 2025Dec 4, 2025Jan 25, 2026Mar 18, 2026May 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
GPT-5 highInput Modality=Text, LLM-as-a-Judge=GPT-4o, Evaluation Protocol=Avg@42025.1195
Gemini 2.5 ProInput Modality=Text, LLM-as-a-Judge=GPT-4o, Evaluation Protocol=Avg@42025.1193.3
Claude Sonnet 4.5Input Modality=Text, LLM-as-a-Judge=GPT-4o, Evaluation Protocol=Avg@42025.1175
Uni-OPDDistillation Scenario=Single-Teacher Distillation, Model Scale=4B2026.0563.3
ExOPDDistillation Scenario=Single-Teacher Distillation, Model Scale=4B2026.0562.7
Uni-OPDDistillation Scenario=Multi-Teacher Distillation, Model Scale=4B2026.0562.3
ExOPDDistillation Scenario=Multi-Teacher Distillation, Model Scale=4B2026.0561
OPDDistillation Scenario=Multi-Teacher Distillation, Model Scale=4B2026.0560.9
TeacherTraining Strategy=RL2026.0560.1
ExPODistillation Scenario=Single-Teacher Distillation, Model Scale=4B2026.0558.7
SFTDistillation Scenario=Multi-Teacher Distillation, Model Scale=4B2026.0558.5
OPDDistillation Scenario=Single-Teacher Distillation, Model Scale=4B2026.0557.9
ExPODistillation Scenario=Multi-Teacher Distillation, Model Scale=4B2026.0557.5
Sora-2 AudioInput Modality=Audio, LLM-as-a-Judge=GPT-4o, Evaluation Protocol=Avg@42025.1146.7
Sora-2 Last FrameInput Modality=Last Frame, LLM-as-a-Judge=GPT-4o, Evaluation Protocol=Avg@42025.1138.3
Oat-ZeroBase Model=Qwen2.5-Math-7B2025.0733.4
Prefix-RFTBase Model=Qwen2.5-Math-7B2025.0731.8
GRPOModel=Qwen3-4B-Base, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=true2026.0530.7
LUFFYBase Model=Qwen2.5-Math-7B2025.0729.4
GRPOModel=Qwen3-4B-Base, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=false2026.0529.2
GRPOModel=Qwen3-4B-Base, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=false2026.0528.3
ReLIFTBase Model=Qwen2.5-Math-7B2025.0728.2
GRPOModel=Qwen3-4B-Base, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=true2026.0527.9
GRPOModel=Qwen2.5-Math-7B, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=true2026.0527.8
SimpleRL-ZeroBase Model=Qwen2.5-Math-7B2025.0727
SFT+RFTBase Model=Qwen2.5-Math-7B2025.0725.8
GRPOModel=Qwen2.5-Math-7B, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=true2026.0525.4
RFTBase Model=Qwen2.5-Math-7B2025.0725.1
StudentModel Scale=4B2026.0523
SFTBase Model=Qwen2.5-Math-7B2025.0722.2
GRPOModel=Qwen2.5-Math-7B, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=false2026.0521.3
UFTBase Model=Qwen2.5-Math-7B2025.0720.8
GRPOModel=Qwen2.5-Math-7B, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=false2026.0520.7
RL w/ SFT LossBase Model=Qwen2.5-Math-7B2025.0719.5
GRPOModel=Qwen3-1.7B-Base, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=true2026.0516.5
GRPOModel=Qwen3-1.7B-Base, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=true2026.0516.2
GRPOModel=Qwen3-1.7B-Base, QB (Query-preserved mini batching)=false, RB (Reward-balanced batching)=false2026.0512.4
GRPOModel=Qwen3-1.7B-Base, QB (Query-preserved mini batching)=true, RB (Reward-balanced batching)=false2026.0511.6
Qwen2.5-Math-7B2025.0711.5