List your company
ResearchFormerly WizWand

Instruction Following on IFEval (IFEval Score)

94.64IFEval ScoreGPT-5-2025-08-07
58.884868.167477.4586.7326Sep 29, 2025Nov 14, 2025Dec 31, 2025Feb 15, 2026Apr 3, 2026May 19, 2026Jul 5, 2026
Updated 3mo ago

Evaluation results

MethodLinks
GPT-5-2025-08-07Setting=high2026.0594.64
JTMode=Think2026.0594.64
TeleChatMode=Thinking2026.0594.27
Kimi-K2.52026.0593.9
GLM-5Precision=FP82026.0593.16
Step-3.5-Flash2026.0593.16
o4-mini-2025-04-16Setting=high2026.0592.98
Gemini-3-Pro-Preview2026.0592.79
GPT-5-mini-2025-08-07Setting=high2026.0592.79
Kimi-K2Mode=Thinking2026.0592.42
o3-2025-04-162026.0592.24
o3-mini-2025-04-16Setting=high2026.0591.87
DeepSeek-V3.2-Speciale2026.0591.68
Qwen3.5-9B (released)Model Size=9B, Training Protocol=Released, Decoding Strategy=Native AR decoding2026.0691.31
MiniMax-M2.52026.0591.13
GLM-4-72026.0590.2
gpt-oss-120bSetting=high2026.0590.2
MiniMax-M22026.0590.2
GPT-5-nano-2025-08-07Setting=high2026.0590.2
LongCat-Flash-Chat2026.0590.2
Gemini-2.5-Pro2026.0590.02
Qwen3.5-4B (released)Model Size=4B, Training Protocol=Released, Decoding Strategy=Native AR decoding2026.0690.02
DeepSeek-V3.22026.0589.65
Qwen3Parameters=30B, Active Parameters=A3B, Mode=Thinking, Version=25072026.0589.65
8B-A-GRPOScale=8B2025.0989.65
Mimo-V2-Flash2026.0589.46
Qwen3-NextParameters=80B, Active Parameters=A3B, Mode=Thinking2026.0589.46
gpt-oss-20bSetting=high2026.0588.91
ERNIE-4.5Parameters=300B, Active Parameters=A47B, Type=PT2026.0588.91
GLM-4.62026.0588.72
Finix-P1-32BMode=Thinking2026.0588.72
Qwen3-4BMode=Thinking, Version=25072026.0588.54
Kimi-K2Type=Instruct2026.0588.54
Claude Sonnet 4Mode=Thinking2026.0588.35
Qwen3Parameters=235B, Active Parameters=A22B, Type=Instruct, Version=25072026.0588.35
GPT-4.1-202504142026.0588.17
Qwen3Parameters=235B, Active Parameters=A22B, Mode=Thinking, Version=25072026.0587.8
Qwen3-NextParameters=80B, Active Parameters=A3B, Type=Instruct2026.0587.62
8B-GRPOScale=8B2025.0987.62
Llama4-MaverickParameters=17B, Experts=128E, Type=Instruct2026.0587.06
Qwen3Parameters=235B, Active Parameters=A22B, Mode=Thinking2026.0585.77
BaselinePrecision=BF162026.0785.7
GLM-4.52026.0585.4
GLM-4-32B-04142026.0585.21
AFM-k5984d3s (Ours)Optimization=QAD + Speculative Decoding + SWA2026.0784.5
GLM-Z1-32B-04142026.0584.29
GLM-4.5-Air2026.0584.1
Qwen3Parameters=30B, Active Parameters=A3B, Type=Instruct, Version=25072026.0583.92
Seed-OSS-36BType=Instruct2026.0583.73
Qwen3-4BType=Instruct, Version=25072026.0582.44
Qwen3-8BScale=8B2025.0982.26
DeepSeek-Chat-V3-03242026.0581.89
QwQ-32B2026.0581.52
ERNIE-4.5Parameters=21B, Active Parameters=A3B, Mode=Thinking2026.0581.15
Gemma-3-27B-it2026.0580.96
DeepSeek-R1-05282026.0580.04
Qwen2.5-14B-InstructBase Model=Qwen2.5-14B-Instruct, # Bits=BF16, Decoding Strategy=greedy2026.0579.85
Qwen3.5-2B (released)Model Size=2B, Training Protocol=Released, Decoding Strategy=Native AR decoding2026.0679.48
doubao-seed-1.6Mode=thinking, Version=2506152026.0578.93
FlexRound+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0578
FlexRoundBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0577.82
FlexRound+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0577.08
MiniCPM-Sala2026.0576.89
Qwen3.5-9B + AR-SFTModel Size=9B, Training Protocol=AR-SFT, Decoding Strategy=Native AR decoding2026.0676.52
OmniQuant+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0575.42
OmniQuant+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0575.23
FlexRoundBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0575.05
OmniQuantBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0574.31
OmniQuantBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0573.94
FLARE-4BModel Size=4B, Training Protocol=FLARE, Decoding Strategy=AR-Trust sampling2026.0673.2
Block-AP+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0572.46
Qwen3.5-4B + AR-SFTModel Size=4B, Training Protocol=AR-SFT, Decoding Strategy=Native AR decoding2026.0672.46
Block-AP+LFQBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0572.27
Block-APBase Model=Qwen2.5-14B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0571.72
FlexRound+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0571.35
FLARE-9BModel Size=9B, Training Protocol=FLARE, Decoding Strategy=AR-Trust sampling2026.0671.35
Qwen2.5-7B-InstructBase Model=Qwen2.5-7B-Instruct, # Bits=BF16, Decoding Strategy=greedy2026.0570.79
Block-APBase Model=Qwen2.5-14B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0570.79
FlexRoundBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0569.5
OmniQuant+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0569.5
FLARE-2BModel Size=2B, Training Protocol=FLARE, Decoding Strategy=AR-Trust sampling2026.0668.95
OmniQuant+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0568.58
OmniQuantBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0568.21
OmniQuantBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0568.21
Block-AP+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0568.02
FlexRound+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0567.84
Qwen3.5-2B + AR-SFTModel Size=2B, Training Protocol=AR-SFT, Decoding Strategy=Native AR decoding2026.0667.65
Block-APBase Model=Qwen2.5-7B-Instruct, # Bits=W4, Decoding Strategy=greedy2026.0566.73
FlexRoundBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0566.54
Block-AP+LFQBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0563.77
Block-APBase Model=Qwen2.5-7B-Instruct, # Bits=W3g128, Decoding Strategy=greedy2026.0561
Hunyuan-A13BType=Instruct2026.0560.26