List your company
ResearchFormerly WizWand

Instruction Following on IFBench (Accuracy %)

76.5AccuracyQwen3.5-397B-A17B
24.53851.565Jun 15, 2026Jun 17, 2026Jun 20, 2026Jun 23, 2026Jun 26, 2026Jun 29, 2026Jul 2, 2026
Updated 3mo ago

Evaluation results

MethodLinks
Qwen3.5-397B-A17BParams=397B2026.0676.5
GLM-5Params=744B2026.0676.5
MiniMax M2.7Params=229B2026.0676
Gemma 4 31BModel size=31B, Thinking mode=true2026.0776
VibeThinker-3BParams=3B2026.0674.5
VibeThinker-3BParams=3B2026.0674.5
Qwen3.6 PlusParams=N/A2026.0674.2
Gemma 4 12BModel size=12B, Thinking mode=true2026.0774
GPT-5 (high)Params=N/A2026.0673.1
Gemma 4 26B-A4BModel size=26B-A4B, Thinking mode=true2026.0772
Nemotron-3-NanoParams=30B2026.0671.5
Gemini 3 ProParams=N/A2026.0670.4
Kimi K2.5Params=1T2026.0670
GPT-OSS (high)Params=120B2026.0669.5
OpenAI o3 (high)Params=N/A2026.0669.3
TeacherModel=MiMo-V2-Flash, Status=RL-trained2026.0668.9
MOPDModel=MiMo-V2-Flash2026.0666.7
GPT-OSS-20B (high)Params=20B2026.0665
MiMo v2 FlashParams=309B2026.0664.2
DeepSeek V3.2Params=671B2026.0660.7
Qwen3.5-4BParams=4B2026.0659.2
Claude Opus 4.5Params=N/A2026.0658
StudentModel=MiMo-V2-Flash2026.0655.4
Grok 4Params=N/A2026.0653.7
Qwen3-4B-Thinking-2507Params=4B2026.0652.9
Phi4-Reasoning-PlusParams=14B2026.0651.7
Qwen3-235B-A22B-ThinkingParams=235B2026.0651.2
Gemini 2.5 ProParams=N/A2026.0648.7
Gemma-4-itParams=12B2026.0645.2
Gemma 4 E4BModel size=E4B, Thinking mode=true2026.0744
DeepSeek R1 0528Params=671B2026.0639.6
Gemma 4 E2BModel size=E2B, Thinking mode=true2026.0738
GLM-4.5-AirParams=106B2026.0637.6
Gemini 2.5 FlashParams=N/A2026.0636.1
Ministral-3-Reasoning-2512Params=14B2026.0632.3
Gemma 3 27BModel size=27B, Thinking mode=false2026.0732
Mimo7B-RL-0530Params=7B2026.0631.6
OpenReasoning-NemotronParams=7B2026.0631.3
Olmo-3-ThinkParams=7B2026.0630
SmolLM3Params=3B2026.0627.6
Hunyuan-4B-InstructParams=4B2026.0626.5