Skip to content

Benchmark

HELM Capabilities: IFEval - IFEval Strict Acc

Instruction following benchmarks measure how closely models follow instructions. Scores are percentages. Higher is better.

Models measured
57
Scores recorded
59
Sources
1
IQ weight
1.66

Score distribution

82.7%
Mean
83.4%
Median
7.3%
Std dev
74.3%
10th percentile
91.8%
90th percentile
56.7%
Min
56.7%95.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.66 for chat models, measured over 57 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Grok 3 Mini· Stanford HELM specialized suites data docs95.1%
2Grok 4· Stanford HELM specialized suites data docs94.9%
3GPT 5.1· Stanford HELM specialized suites data docs93.5%
4GPT-5 nano· Stanford HELM specialized suites data docs93.2%
5o4 Mini· Stanford HELM specialized suites data docs92.9%
6GPT-5 mini· Stanford HELM specialized suites data docs92.7%
7Claude Opus 4setting not named by the source· Stanford HELM specialized suites data docs91.8%
8Llama 4 Maverick· Stanford HELM specialized suites data docs90.8%
9GPT-4.1 mini· Stanford HELM specialized suites data docs90.4%
10gemini-2.5-flash-preview-04-17· Stanford HELM specialized suites data docs89.8%
11Granite 4.0 H Small· Stanford HELM specialized suites data docs89%
12Grok 3· Stanford HELM specialized suites data docs88.4%
13Gemini 3 Pro Preview· Stanford HELM specialized suites data docs87.6%
14Mistral Large2411· Stanford HELM specialized suites data docs87.6%
15GPT 5· Stanford HELM specialized suites data docs87.5%
16OpenAI o3· Stanford HELM specialized suites data docs86.9%
17Claude 3.5 Sonnet· Stanford HELM specialized suites data docs85.6%
18Kimi K2 Instruct· Stanford HELM specialized suites data docs85%
19Claude Sonnet 4.5· Stanford HELM specialized suites data docs85%
20Claude Opus 4thinking, 10k· Stanford HELM specialized suites data docs84.9%
21GPT-4.1 nano· Stanford HELM specialized suites data docs84.3%
22Gemini 2.0 Flash· Stanford HELM specialized suites data docs84.1%
23Claude Sonnet 4thinking, 10k· Stanford HELM specialized suites data docs84%
24Gemini 2.5 Pro Preview 03-25· Stanford HELM specialized suites data docs84%
25Claude Sonnet 4setting not named by the source· Stanford HELM specialized suites data docs83.9%
26GPT 4.1· Stanford HELM specialized suites data docs83.8%
27gemini-1.5-pro-002· Stanford HELM specialized suites data docs83.7%
28GPT OSS 120B· Stanford HELM specialized suites data docs83.6%
29Qwen3 235B A22B Instruct 2507· Stanford HELM specialized suites data docs83.5%
30Claude Sonnet 3.7· Stanford HELM specialized suites data docs83.4%
31DeepSeek V3· Stanford HELM specialized suites data docs83.2%
32gemini-1.5-flash-002· Stanford HELM specialized suites data docs83.1%
33Gemini 2.0 Flash-Lite Preview· Stanford HELM specialized suites data docs82.4%
34Palmyra X5· Stanford HELM specialized suites data docs82.3%
35Meta Llama 3.1 70B Instruct Turbo· Stanford HELM specialized suites data docs82.1%
36Llama 4 Scout· Stanford HELM specialized suites data docs81.8%
37GPT 4o· Stanford HELM specialized suites data docs81.7%
38Nova Prov1· Stanford HELM specialized suites data docs81.5%
39GLM 4.5 Air· Stanford HELM specialized suites data docs81.2%
40Llama 3.1 405B Instruct Turbo· Stanford HELM specialized suites data docs81.1%
41Gemini 2.5 Flash-Lite· Stanford HELM specialized suites data docs81%
42Qwen3 Next 80B A3Bthinking· Stanford HELM specialized suites data docs81%
43Qwen2.5 72B Instruct Turbo· Stanford HELM specialized suites data docs80.6%
44Nova Premier 1.0· Stanford HELM specialized suites data docs80.3%
45Claude Haiku 4.5· Stanford HELM specialized suites data docs80.1%
46Claude Haiku 3.5· Stanford HELM specialized suites data docs79.2%
47DeepSeek R1· Stanford HELM specialized suites data docs78.4%
48GPT-4o mini· Stanford HELM specialized suites data docs78.2%
49Olmo 2 32B Instruct· Stanford HELM specialized suites data docs78%
50Nova Litev1· Stanford HELM specialized suites data docs77.6%

Showing 1–50 of 59 results.

Data credits: Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance