LLM Reference
IFEvalactiveGeneral

IFEval: Instruction-Following Evaluation

Metric: Instruction Following Score (higher is better)Introduced: 2023

IFEval measures LLM ability to follow verifiable formatting instructions (e.g., output length, keyword inclusion, capitalization) with exact-match checking. High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

12

tracked on this benchmark

Score band

94.8 – 38.5

best → lowest tracked

Snapshot trend

+3.87

Apr 29 → Jun 26 · 1 models

Leaderboard

Tracked models ranked by Instruction Following Score (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Agents-A1
Version: IFEval accuracy, vendor-reported on Agents-A1 model cardHarness: Not recordedEvaluator: Not recordedObserved: Jun 26, 2026Confidence: Not recordedSource

Notes: Official Intern Science model card reports Agents-A1 at 94.82 on IFEval. Evaluator/source: Intern Science self-reported model-card/project benchmark table. Public harness details are not fully reproducible from the seed row. Confidence: medium for reported value, low-medium for reproducibility. Recommended seed value: 94.82.

94.8
2
Qwen3.5-397B-A17B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 19, 2026Confidence: Not recordedSource
92.6
3
GPT-5.5
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 29, 2026Confidence: Not recordedSource
92.1
4
LFM2.5 8B A1B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
91.8
5
Kimi K2.6
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 29, 2026Confidence: Not recordedSource
89.8
6
Llama 3 70B Instruct
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
77.8
7
Gemma 2 9B Instruct
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
65.5
8
Llama 3 8B Instruct
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
59.5
9
Qwen2-7B-Instruct
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
57.8
10
Phi-3 Mini 4k
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
45.0
11
Gemma 7B Instruct
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
42.6
12
Mistral 7B Instruct v0.3
Version: v2Harness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
38.5

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the IFEval benchmark measure?

IFEval measures LLM ability to follow verifiable formatting instructions (e.g., output length, keyword inclusion, capitalization) with exact-match checking. On this page it lists 12 tracked model variants where higher is better.

Is a higher Instruction-Following Evaluation score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this Instruction-Following Evaluation data?

This benchmark was last reviewed on Apr 15, 2026. The tracked score average moved +3.87 points across the last 3 snapshots.

Related benchmarks

Last reviewed: Apr 15, 2026