Tests explained · People's votes

People's votes on following instructions

People compared two anonymous answers to requests with specific instructions.

Run by

How to read it

A rating from thousands of head-to-head votes. Higher means people preferred its answers more often. The thin line is the likely range.

Used for

6 tasks on this site.

How the models did

Rating from blind votes · Higher is better

Claude Opus 5.5 and Claude Fable 5.1 are neck and neck at the top, followed by Gemini 3.8 Flash and Kimi K3.

Claude Opus 5.5 · Anthropic1st

Claude Fable 5.1 · Anthropic2nd

Gemini 3.8 Flash · Google DeepMind3rd

Kimi K3 · Moonshot AI4th

GPT-5.6 Sol · OpenAI5th

DeepSeek V4.1 Flash · DeepSeek6th

Gemini 3.6 Flash · Google DeepMind7th

Gemini 3.1 Pro · Google DeepMind8th

GPT-6 Astra · OpenAI9th

Claude Sonnet 5 · Anthropic10th

DeepSeek V4 Pro · DeepSeek11th

GPT-5.6 Luna · OpenAI12th

Gemini 3.5 Flash-Lite · Google DeepMind13th

Claude Haiku 4.5 · Anthropic14th

GPT-6 Luna · OpenAINot tested yet

GPT-6 Sol · OpenAINot tested yet

Source: LMArena · votes as of 25 Sept 2026

Where we use it

Each task page labels this test by how closely it matches the task.

Task	How close
Best AI for writing	Tests a related skill
Best AI for essays and academic writing	Tests a related skill
Best AI for teachers	Tests a related skill
Best AI for CVs, job applications and interviews	Tests a related skill
Best AI for notes and meeting notes	Tests a related skill
Best AI for presentations and PowerPoint	Tests a related skill