Inference
EN
Jeffrey Zhou et al.

Instruction-Following Eval

A benchmark of 541 prompts built from 25 verifiable instruction types. It tests whether a model follows checkable constraints such as keyword, length, casing, and response-format requirements.

ModelsModels with a result.70
Top resultBest result on this benchmark.96.8%LFM2.5-24B-A2B
Top-3 spreadPoints from first to third.1.7 pts
YearYear of release.2023

Result and price

26 of 70 results

Results

70 results
#ModelResultThe score from the source.
01LFM2.5-24B-A2BAbove fitFar above this model's index.96.8%
02LFM3-1.7B96.7%
03Flint95.1%
04Qwen3.5-27B95.0%
05Agents-A194.8%
05Agents-A1-4B94.8%
07Qwen3.7 Plus94.6%
08Qwen3.7 Max94.3%
08Qwen3.6 Plus94.3%
10dots3-note Preview93.9%
10Kimi K2.593.9%
10o3-mini93.9%
13Spark-X2.6-4B93.6%
14Qwen3.5-122B-A10B93.4%
15Claude 3.7 Sonnet Thinking93.2%
16Qwen3.5 397B92.6%
16GLM-592.6%
18K-EXAONE 2.092.4%
19o192.2%
20Llama 3.3 70BAbove fitFar above this model's index.92.1%
21Qwen3.5-35B-A3B91.9%
22LFM2.5-8B-A1BAbove fitFar above this model's index.91.8%
23Ternary Bonsai 2 27B91.3%
24Claude Opus 4.590.9%
25Claude 3.7 Sonnet90.8%
26Mellum2.1-12B-A2.5B-Thinking90.6%
27Kimi K289.8%
28Nemotron Ultra 253B89.5%
29Qwen3 235B 250788.7%
30Llama 3.1 405B88.6%
31GPT-4.1 mini88.5%
32Qwen3 235B 2507 (Reasoning)87.8%
33Llama 3.1 70B87.5%
34GPT-4.187.4%
35MiniCPM5-2B86.7%
36DeepSeek V386.1%
37Claude 3.5 Haiku85.9%
38ZAYA1-8B85.6%
39Qwen3 32B (Reasoning)85.0%
40Phi-4-reasoning-plus84.9%
41Sarvam 105B84.8%
42QwQ-32B83.9%
43Phi-4-reasoning83.4%
43Qwen3 235B-A22B (Reasoning)83.4%
45DeepSeek-R183.3%
46Qwen3 235B-A22B83.2%
46GPT-4.1 nano83.2%
46Qwen3 32B83.2%
49OrcaCyber-Nano-1.082.8%
50LFM2.5-VL-3B82.3%
51Ternary Bonsai 8B81.8%
52Kanana-2 3B Instruct81.0%
53Celeris-180.8%
54MiniCPM5-1B80.4%
54Llama 3.1 8B80.4%
561-bit Bonsai 8B79.8%
57Granite-4.0-H-1B78.5%
58Kanana-2 1.3B Instruct77.6%
59Granite-4.0-1B77.4%
60LFM2.5-350M77.0%
61Mellum2-12B-A2.5B-Thinking76.5%
62Mellum2-12B-A2.5B-Instruct75.8%
63Ternary Bonsai 4B72.1%
64LFM2.5-230M71.7%
65Ternary Bonsai 1.7B70.1%
661-bit Bonsai 4B69.6%
671-bit Bonsai 1.7B63.0%
68Granite-4.0-H-350M61.6%
69LFM2.5-VL-450M61.2%
70Granite-4.0-350M55.4%