Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Instruction Following benchmark
AgentIF: instruction following in agentic scenarios. Every model the catalog carries a published AgentIF value for, ranked by that value.
Instruction-following benchmark built from real agentic applications, with long instructions that carry many constraints. 2 models in the ModelChoice catalog have a published value.
Each point is a model with a published AgentIF value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
2 models with a published AgentIF value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Instruction success rate |
|---|---|---|---|
| 1 | NVIDIA | 25.81 | |
| 2 | NVIDIA | 22.95 |
Evidence key: Observed
Rows are ordered by the value AgentIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
What the benchmark measures and how its value is published, from the source’s own record.
AgentIF (AgentIF: instruction following in agentic scenarios) is an instruction following benchmark published by AgentIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios. Instruction-following benchmark built from real agentic applications, with long instructions that carry many constraints. Its evaluation set is described by the source as agentic system prompts and instructions. The source rates its difficulty as complex agentic instructions.
Each model's result is published by AgentIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios as instruction success rate. ModelChoice prints that value to 2 decimal places, exactly as published, and never converts or rescales it. This page ranks the 2 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
All leaderboards · Instruction Following capability leaderboard