List your company
ResearchFormerly WizWand

Instruction Following on AlpacaEval LC 2

80.9Win RateQwen 3 VL 32B Think
-1.57219.83941.2562.661Oct 10, 2025Nov 15, 2025Dec 21, 2025Jan 26, 2026Mar 3, 2026Apr 8, 2026May 14, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen 3 VL 32B ThinkModel Family=Qwen 3, Parameter Count=32B, Thinking Capability=true2025.1280.9
Olmo 3 Think (DPO)Training Stage=DPO, Model Family=Olmo 3, Parameter Count=32B, Thinking Capability=true2025.1278.6
Qwen 3 32BModel Family=Qwen 3, Parameter Count=32B, Thinking Capability=false2025.1275.6
Olmo 3 Think (Final 3.0)Training Stage=Final Think 3.0, Model Family=Olmo 3, Parameter Count=32B, Thinking Capability=true2025.1274.2
Multi-DPOPModel=Qwen3-30B2025.1069.7
Olmo 3 Think (SFT)Training Stage=SFT, Model Family=Olmo 3, Parameter Count=32B, Thinking Capability=true2025.1269.1
Olmo 3.1 Think 32BTraining Stage=Final Think 3.1, Model Family=Olmo 3.1, Parameter Count=32B, Thinking Capability=true2025.1269.1
GRPOModel=Qwen3-30B2025.1068.9
DPOModel=Qwen3-30B2025.1068.5
BaseModel=Qwen3-30B2025.1067.4
SFTModel=Qwen3-30B2025.1065.3
Multi-DPOPModel=GLM-4-9B2025.1064.2
GRPOModel=GLM-4-9B2025.1063.8
DPOModel=GLM-4-9B2025.1063.5
BaseModel=GLM-4-9B2025.1062.3
SFTModel=GLM-4-9B2025.1060.7
Multi-DPOPModel=Qwen3-4B2025.1058.4
GRPOModel=Qwen3-4B2025.1057.5
DPOModel=Qwen3-4B2025.1057.2
BaseModel=Qwen3-4B2025.1056.2
SFTModel=Qwen3-4B2025.1054.5
Qwen3-4B (Non-thinking)Variant=Non-thinking2026.0553.5
Qwen3-4B-GRLO+RLVRMethod=GRLO+RLVR2026.0551.1
Qwen3-4B-GRLOMethod=GRLO2026.0545.8
Multi-DPOPModel=Gemma-3-4B2025.1043.1
GRPOModel=Gemma-3-4B2025.1042.8
DPOModel=Gemma-3-4B2025.1042.5
BaseModel=Gemma-3-4B2025.1041.2
SFTModel=Gemma-3-4B2025.1039.8
PROSPER-JCBackbone=Qwen2.5-7B-Instruct, Model Scale=7B, LLM Judge=gpt-5-mini2026.0238.21
PROSPERBackbone=Qwen2.5-7B-Instruct, Model Scale=7B, LLM Judge=gpt-5-mini2026.0237.61
Multi-DPOPModel=Tulu-3.1-8B2025.1036.9
DPOModel=Tulu-3.1-8B2025.1036.7
GRPOModel=Tulu-3.1-8B2025.1036.3
BaseModel=Tulu-3.1-8B2025.1034.5
Llama3.2-3B-GRLO+RLVRBackbone=Llama3.2-3B, Training Strategy=GRLO+RLVR2026.0533.9
PROSPER-VBBackbone=Qwen2.5-7B-Instruct, Model Scale=7B, LLM Judge=gpt-5-mini2026.0233.64
SFTModel=Tulu-3.1-8B2025.1033.1
Llama3.2-3B-GRLOBackbone=Llama3.2-3B, Training Strategy=GRLO2026.0531.5
General-Reasoner-4B2026.0530.8
DS-R1 32BModel Family=DeepSeek-R1, Parameter Count=32B, Thinking Capability=true2025.1226.2
Qwen2.5-7B-InstructModel Scale=7B, LLM Judge=gpt-5-mini2026.0225.32
Llama3.2-3B-InstructBackbone=Llama3.2-3B, Training Strategy=Instruct2026.0522.9
RLCFBackbone=Qwen2.5-7B-Instruct, Model Scale=7B, LLM Judge=gpt-5-mini2026.0217.24
Llama3.2-3B-SFTBackbone=Llama3.2-3B, Training Strategy=SFT2026.058.5
Llama3.2-3B-RLVRBackbone=Llama3.2-3B, Training Strategy=RLVR2026.058.1
Qwen3-4B-OpenSFTVariant=OpenSFT2026.057.6
Qwen3-4B-MathSFTVariant=MathSFT2026.057.3
Qwen3-4B-BaseBackbone=Qwen3-4B2026.051.6