reasoning benchmark
Multi-IF
Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.
Updated Aug 7, 2026
Published models20
Registry coverage20
MetricScore
EvidenceB
Multi-IF leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
Multi-IF highlights
The top published results on this benchmark's own scale.
What is Multi-IF?
Definition and scoring fields from the benchmark registry.
Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- Multi-IF
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- multi-if|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about Multi-IF.
Which model scores highest on Multi-IF?
Qwen3-235B-A22B-Thinking-2507 is currently ranked first with 80.6%.
What does Multi-IF measure?
Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
20 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.