benchgap
Instruction following

IFEval leaderboard

As of 2026-11-22, the highest measured score on IFEval is 96.8% by LFM2.5-24B-A2B. 137 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1LFM2.5-24B-A2B96.8%measured
2LFM3-1.7B96.7%measured
3MAI-Thinking-195.3%estimated ± 3.6 pp, high confidence
4Flint95.1%measured
5Qwen3.8 Max95.0%estimated ± 3.6 pp, high confidence
6Qwen3.5-27B95.0%measured
7Inkling-Small95.0%estimated ± 3.6 pp, high confidence
8Sage-A1.594.9%estimated ± 3.6 pp, high confidence
9Qwen3.8-Omni-Flash94.9%estimated ± 3.6 pp, high confidence
10Qwen3.8-Flash-Next94.8%estimated ± 3.6 pp, high confidence
11Agents-A194.8%measured
12Agents-A1-4B94.8%measured
13Beam 1.0 501B-A23B FP894.8%estimated ± 3.6 pp, high confidence
14Solar Open 294.7%estimated ± 3.6 pp, high confidence
15Inkling94.6%estimated ± 3.6 pp, high confidence
16Beam 1.0 501B-A23B94.6%estimated ± 3.6 pp, high confidence
17Qwen3.7 Plus94.6%measured
18Qwen3.8-27B94.6%estimated ± 3.6 pp, high confidence
19Granite 4.2 8B94.6%estimated ± 3.6 pp, high confidence
20Beam 1.0 501B-A23B NVFP494.5%estimated ± 3.6 pp, high confidence
21Ripple-65B Preview94.5%estimated ± 3.6 pp, high confidence
22Qwen3.6 Plus94.3%measured
23Qwen3.7 Max94.3%measured
24Granite 4.2 30B94.2%estimated ± 3.6 pp, high confidence
25Mercury 2.594.2%estimated ± 3.6 pp, high confidence
26Muse Glimmer 30B94.2%estimated ± 3.6 pp, high confidence
27Mercury 394.1%estimated ± 3.6 pp, high confidence
28A.X K294.0%estimated ± 3.6 pp, high confidence
29dots3-note Preview93.9%measured
30Kimi K2.593.9%measured
31o3-mini93.9%measured
32Ling 3.0 Flash93.7%estimated ± 3.6 pp, high confidence
33Granite 4.2 3B93.6%estimated ± 3.6 pp, high confidence
34Spark-X2.6-4B93.6%measured
35Ling 3.0 Flash FP893.4%estimated ± 3.6 pp, high confidence
36Qwen3.5-122B-A10B93.4%measured
37Nemotron 3.5 Lightning 30B A3B NVFP493.3%estimated ± 3.6 pp, high confidence
38MiniMax M393.3%estimated ± 2.8 pp, medium confidence
39Nemotron 3 Ultra93.2%estimated ± 2.8 pp, medium confidence
40Grok 4.393.2%estimated ± 2.8 pp, medium confidence
41Claude 3.7 Sonnet Thinking93.2%measured
42MiMo-V2.5-Pro93.2%estimated ± 2.8 pp, high confidence
43GPT-5.2-Codex93.1%estimated ± 2.8 pp, high confidence
44Gemini 3.1 Pro93.1%estimated ± 2.8 pp, high confidence
45Qwen 3.6 Max (preview)93.1%estimated ± 2.8 pp, high confidence
46DeepSeek V4 Pro93.1%estimated ± 2.8 pp, high confidence
47Gemini 3.5 Flash93.1%estimated ± 2.8 pp, high confidence
48GLM-5.193.1%estimated ± 2.8 pp, high confidence
49Kimi K2.693.0%estimated ± 2.8 pp, high confidence
50GPT-5.4 nano93.0%estimated ± 2.8 pp, high confidence
51GPT-5.593.0%estimated ± 2.8 pp, high confidence
52Muse Spark93.0%estimated ± 2.8 pp, high confidence
53MiniMax M2.793.0%estimated ± 2.8 pp, high confidence
54Gemma 4 31B93.0%estimated ± 2.8 pp, high confidence
55GPT-5.293.0%estimated ± 2.8 pp, high confidence
56GPT-5.3 Codex93.0%estimated ± 2.8 pp, high confidence
57Command A+93.0%estimated ± 2.8 pp, high confidence
58GPT-5.493.0%estimated ± 2.8 pp, high confidence
59Nemotron 3.5 Super VL 120B A12B NVFP493.0%estimated ± 3.6 pp, high confidence
60Gemma 4 12B93.0%estimated ± 2.8 pp, high confidence
61GLM-5.292.9%estimated ± 2.8 pp, high confidence
62GPT-5.4 mini92.9%estimated ± 2.8 pp, high confidence
63GLM-5-Turbo92.9%estimated ± 2.8 pp, high confidence
64GPT-592.9%estimated ± 2.8 pp, high confidence
65GPT-5.192.9%estimated ± 2.8 pp, high confidence
66GPT-5.6 Sol92.9%estimated ± 2.8 pp, high confidence
67Gemma 4 26B A4B92.9%estimated ± 2.8 pp, high confidence
68o392.9%estimated ± 2.8 pp, high confidence
69GPT-5.6 Terra92.9%estimated ± 2.8 pp, high confidence
70Solar Pro 392.9%estimated ± 2.8 pp, high confidence
71Gemini 3 Pro92.8%estimated ± 2.8 pp, high confidence
72GPT-5.1-Codex92.8%estimated ± 2.8 pp, high confidence
73GPT-5.1-Codex-Max92.8%estimated ± 2.8 pp, high confidence
74Nemotron 3.5 Super VL 120B A12B92.8%estimated ± 3.6 pp, high confidence
75GPT-OSS 120B92.8%estimated ± 2.8 pp, high confidence
76MiMo-V2-Pro92.8%estimated ± 2.8 pp, high confidence
77Mistral Medium 3.5 128B92.8%estimated ± 2.8 pp, high confidence
78o4-mini92.7%estimated ± 2.8 pp, high confidence
79GLM-4.792.7%estimated ± 2.8 pp, high confidence
80Qwen3.6-27B92.7%estimated ± 2.8 pp, high confidence
81Step 3.7 Flash92.7%estimated ± 2.8 pp, high confidence
82GLM-592.6%measured
83Qwen3.5 397B92.6%measured
84GPT-OSS 20B92.5%estimated ± 2.8 pp, high confidence
85K-Exaone92.5%estimated ± 2.8 pp, high confidence
86Qwen3.6-35B-A3B92.5%estimated ± 2.8 pp, high confidence
87Claude Fable 592.4%estimated ± 2.8 pp, high confidence
88Nemotron 3 Nano Omni 30B A3B92.4%estimated ± 2.8 pp, high confidence
89Kimi K2.7 Code92.4%estimated ± 2.8 pp, high confidence
90K-EXAONE 2.092.4%measured
91Claude Opus 4.892.4%estimated ± 2.8 pp, high confidence
92GLM-5V-Turbo92.3%estimated ± 2.8 pp, high confidence
93Nemotron 3.5 Bolt E4B92.3%estimated ± 2.8 pp, high confidence
94o192.2%measured
95Llama 3.3 70B92.1%measured
96Claude Opus 4.792.1%estimated ± 2.8 pp, high confidence
97Claude Opus 4.5 Thinking92.0%estimated ± 2.8 pp, high confidence
98North Mini Code92.0%estimated ± 2.8 pp, high confidence
99Ling 2.6 Flash92.0%estimated ± 2.8 pp, high confidence
100Qwen3.5-35B-A3B91.9%measured
101Trinity-Large-Preview91.9%estimated ± 2.8 pp, high confidence
102Trinity-Large-Thinking91.9%estimated ± 2.8 pp, high confidence
103LFM2.5-8B-A1B91.8%measured
104Gemini 3 Flash91.7%estimated ± 2.8 pp, high confidence
105Claude 4 Opus Thinking91.6%estimated ± 2.8 pp, high confidence
106Grok 491.6%estimated ± 2.8 pp, high confidence
107MiMo-V2-Omni91.6%estimated ± 2.8 pp, high confidence
108Claude Opus 4.6 (Adaptive)91.5%estimated ± 2.8 pp, high confidence
109Grok 4.1 Fast (Reasoning)91.5%estimated ± 2.8 pp, high confidence
110Qwen3.5-4B91.4%estimated ± 2.8 pp, high confidence
111Ternary Bonsai 2 27B91.3%measured
112Grok 4 Fast (Reasoning)91.2%estimated ± 2.8 pp, high confidence
113DeepSeek V3.290.9%estimated ± 2.8 pp, high confidence
114Claude Opus 4.590.9%measured
115Gemini 2.5 Pro90.9%estimated ± 2.8 pp, high confidence
116Mistral Small 490.8%estimated ± 2.8 pp, high confidence
117Claude 3.7 Sonnet90.8%measured
118Ultravox v0.6 Llama 3.3 70B90.6%estimated ± 2.8 pp, high confidence
119Mellum2.1-12B-A2.5B-Thinking90.6%measured
120Claude 4 Sonnet90.3%estimated ± 2.8 pp, high confidence
121Hy3 Preview90.1%estimated ± 3.6 pp, high confidence
122Claude Opus 4.690.1%estimated ± 2.8 pp, high confidence
123Gemma 4 E4B90.0%estimated ± 2.8 pp, high confidence
124Qwen3 Max90.0%estimated ± 2.8 pp, high confidence
125Qwen3-Omni-30B-A3B-Thinking89.8%estimated ± 2.8 pp, high confidence
126Claude 4 Opus89.8%estimated ± 2.8 pp, high confidence
127Kimi K289.8%measured
128Llama 4 Maverick89.7%estimated ± 2.8 pp, high confidence
129Nemotron Ultra 253B89.5%measured
130Grok Code Fast 189.3%estimated ± 2.8 pp, high confidence
131Claude Sonnet 4.689.2%estimated ± 2.8 pp, high confidence
132DeepSeek V3 032489.2%estimated ± 2.8 pp, high confidence
133Gemini 2.0 Flash88.9%estimated ± 2.8 pp, high confidence
134MiMo-V2-Flash88.8%estimated ± 2.8 pp, high confidence
135Llama 4 Scout88.7%estimated ± 2.8 pp, high confidence
136Qwen3 235B 250788.7%measured
137Mistral Medium 388.6%estimated ± 2.8 pp, high confidence
138Llama 3.1 405B88.6%measured
139Gemini 2.5 Flash88.5%estimated ± 2.8 pp, high confidence
140GPT-4.1 mini88.5%measured
141LFM2.5-2.6B88.3%estimated ± 3.6 pp, high confidence
142Nova Pro88.2%estimated ± 2.8 pp, high confidence
143Gemma 4 E2B88.2%estimated ± 2.8 pp, high confidence
144DeepSeek V3.188.1%estimated ± 2.8 pp, high confidence
145GLM-4.5-Air88.0%estimated ± 2.8 pp, high confidence
146Nemotron 3 Nano 30B88.0%estimated ± 2.8 pp, high confidence
147GLM-4.687.7%estimated ± 2.8 pp, high confidence
148Grok 4.1 Fast87.6%estimated ± 2.8 pp, high confidence
149Llama 3.1 70B87.5%measured
150Mistral Large 387.4%estimated ± 2.8 pp, high confidence
151GPT-4.187.4%measured
152Claude 3 Haiku87.4%estimated ± 2.8 pp, high confidence
153MiniCPM5-2B86.7%measured
154Devstral Small86.7%estimated ± 2.8 pp, high confidence
155GPT-4o86.5%estimated ± 2.8 pp, high confidence
156Solar Pro 286.2%estimated ± 2.8 pp, high confidence
157DeepSeek V386.1%measured
158Exaone 4.0 32B86.1%estimated ± 2.8 pp, high confidence
159Claude 3.5 Haiku85.9%measured
160LFM2.5-VL-1.6B-Extract85.8%estimated ± 2.8 pp, high confidence
161ZAYA1-8B85.6%measured
162Gemma 3 27B85.0%estimated ± 2.8 pp, high confidence
163Qwen3 32B (Reasoning)85.0%measured
164Phi-4-reasoning-plus84.9%measured
165Gemini 2.5 Flash-Lite84.8%estimated ± 2.8 pp, high confidence
166Sarvam 105B84.8%measured
167Mistral Large 284.6%estimated ± 2.8 pp, high confidence
168Qwen3-Omni-30B-A3B-Instruct84.6%estimated ± 2.8 pp, high confidence
169GPT-4o mini84.5%estimated ± 2.8 pp, high confidence
170QwQ-32B83.9%measured
171Mistral Small 3.183.6%estimated ± 2.8 pp, high confidence
172Phi-4-reasoning83.4%measured
173Qwen3 235B-A22B (Reasoning)83.4%measured
174DeepSeek-R183.3%measured
175GPT-4.1 nano83.2%measured
176Qwen3 235B-A22B83.2%measured
177Qwen3 32B83.2%measured
178OrcaCyber-Nano-1.082.8%measured
179LFM2.5-VL-3B82.3%measured
180Ternary Bonsai 8B81.8%measured
181Kanana-2 3B Instruct81.0%measured
182Celeris-180.8%measured
183MiniCPM5-1B80.4%measured
184Llama 3.1 8B80.4%measured
185Sarvam 30B80.3%estimated ± 2.8 pp, high confidence
1861-bit Bonsai 8B79.8%measured
187Exaone 4.0 1.2B78.8%estimated ± 2.8 pp, high confidence
188Magistral Medium78.6%estimated ± 2.8 pp, high confidence
189Granite-4.0-H-1B78.5%measured
190Magistral Small78.1%estimated ± 2.8 pp, high confidence
191LLaDA2.2-mini77.7%estimated ± 3.6 pp, medium confidence
192Kanana-2 1.3B Instruct77.6%measured
193Granite-4.0-1B77.4%measured
194LFM2.5-350M77.0%measured
195Mellum2-12B-A2.5B-Thinking76.5%measured
196Phi-476.2%estimated ± 2.8 pp, high confidence
197Mellum2-12B-A2.5B-Instruct75.8%measured
198DeepSeek R1 Distill Qwen 32B75.2%estimated ± 2.8 pp, high confidence
199Ternary Bonsai 4B72.1%measured
200LFM2.5-230M71.7%measured
201Ternary Bonsai 1.7B70.1%measured
2021-bit Bonsai 4B69.6%measured
2031-bit Bonsai 1.7B63.0%measured
204Granite-4.0-H-350M61.6%measured
205LFM2.5-VL-450M61.2%measured
206Granite-4.0-350M55.4%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General