Instruction following benchmarks
See how reliably models follow formatting rules, explicit constraints, and multi-step instructions. These evaluations focus on satisfying the requested requirements as well as answering the task.
Published
Updated:
Models tested: 11
Conducting multi-turn conversations with human users is a common yet challenging task for LLMs. It is challenging because conducting multi-turn conversations requires not only…
Top models · scale
Latest result per model · Version not reported
- 1GPT-6.1 Sol81.95 source score
- 2Muse Spark 1.175.3 source score
- 3Gemini 3.1 Pro Preview71.37 source score
Updated:
Models tested: 4
The Precise Instruction Following Prompts Dataset is composed of 1,054 instruction following prompts aimed at assessing the ability of AI models to interpret and execute detailed…
Top models · scale
Latest result per model · Version not reported
- 1DeepSeek R187.75 source score
- 2GPT-4 Turbo Preview83.19 source score
- 3Claude 3 Opus80.12 source score
Updated:
Models tested: 151
Human pairwise preferences on instruction following prompts in Arena’s text evaluation. The original organizer publishes model ratings and uncertainty; these are preference…
Top models · Arena
Latest result per model · Version not reported
- 1Claude Opus 5.5High1,537.02
- 2Gemini 4 ArgonHigh1,533.92
- 3Claude Opus 4.6High1,521.42
Updated:
Models tested: 151
Human pairwise preferences on instruction following prompts in Arena’s text style control evaluation. The original organizer publishes model ratings and uncertainty, adjusting for…
Top models · Arena
Latest result per model · Version not reported
- 1Gemini 4 ArgonHigh1,525.35
- 2Claude Opus 5.5High1,519.46
- 3Claude Opus 4.6High1,511.21
Methodology and provenance
Benchize collects and structures public benchmark reports. Scores, costs, timings, ranks and uncertainty remain attributed to their original sources. Different versions, configurations and measurement units are separate results. Missing fields stay unavailable.