Instruction Following on IFBench (Accuracy %) benchmark leaderboard · SOTA2 Research

Instruction Following on IFBench (Accuracy %)

76.5AccuracyQwen3.5-397B-A17B

Evaluation Results

Method	Links
Qwen3.5-397B-A17B2026.06		76.5
GLM-52026.06		76.5
MiniMax M2.72026.06		76
Gemma 4 31B2026.07		76
VibeThinker-3B2026.06		74.5
VibeThinker-3B2026.06		74.5
Qwen3.6 Plus2026.06		74.2
Gemma 4 12B2026.07		74
GPT-5 (high)2026.06		73.1
Gemma 4 26B-A4B2026.07		72
Nemotron-3-Nano2026.06		71.5
Gemini 3 Pro2026.06		70.4
Kimi K2.52026.06		70
GPT-OSS (high)2026.06		69.5
OpenAI o3 (high)2026.06		69.3
Teacher2026.06		68.9
MOPD2026.06		66.7
GPT-OSS-20B (high)2026.06		65
MiMo v2 Flash2026.06		64.2
DeepSeek V3.22026.06		60.7
Qwen3.5-4B2026.06		59.2
Claude Opus 4.52026.06		58
Student2026.06		55.4
Grok 42026.06		53.7
Qwen3-4B-Thinking-25072026.06		52.9
Phi4-Reasoning-Plus2026.06		51.7
Qwen3-235B-A22B-Thinking2026.06		51.2
Gemini 2.5 Pro2026.06		48.7
Gemma-4-it2026.06		45.2
Gemma 4 E4B2026.07		44
DeepSeek R1 05282026.06		39.6
Gemma 4 E2B2026.07		38
GLM-4.5-Air2026.06		37.6
Gemini 2.5 Flash2026.06		36.1
Ministral-3-Reasoning-25122026.06		32.3
Gemma 3 27B2026.07		32
Mimo7B-RL-05302026.06		31.6
OpenReasoning-Nemotron2026.06		31.3
Olmo-3-Think2026.06		30
SmolLM32026.06		27.6
Hunyuan-4B-Instruct2026.06		26.5