Arena Elo: what the benchmark actually measures
A practical guide to arena-elo, using data from the AI & LLM Index. The goal is a shortlist and a decision, not a tour of every option.
What Arena Elo measures
Arena Elo is a preference benchmark. Human preference rating from pairwise model comparisons.
The index scores all 600 models on a 800–1600 scale. The average is 1349.9 and the top score is 1501.9.
How to read the scores
A benchmark score is a compressed signal. It tells you how a model performed on a specific test distribution, not how it will perform on your data. Read the leaderboard as a way to form a shortlist, then validate the shortlist with your own examples.
On Arena Elo, the gap between the leader and the median is 152.0 points. That gap is the useful part of the chart: it shows how much separation the benchmark creates.
Leaders
Claude Opus 4.5 Max leads at 1501.9, followed by GPT-5.2 Reasoning at 1500.4 and GPT-5.2 Turbo at 1500.3.
Limitations
- Scores in this index are modelled estimates, not official results.
- Benchmark contamination can inflate scores for models trained on similar data.
- A high score does not guarantee good performance on a different task shape.
- Prompt format and evaluation harness can move results by several points.
Top 5 on Arena Elo
| Model | Provider | Score | Input / 1M |
|---|---|---|---|
| Claude Opus 4.5 Max | Anthropic | 1501.9 | $20.90 |
| GPT-5.2 Reasoning | OpenAI | 1500.4 | $20.98 |
| GPT-5.2 Turbo | OpenAI | 1500.3 | $20.73 |
| Claude Opus 4.5 Reasoning | Anthropic | 1497.5 | $20.91 |
| GPT-5.2 Max | OpenAI | 1495.3 | $21.39 |
Related
FAQ
Who leads Arena Elo?
Claude Opus 4.5 Max leads this index at 1501.9.
What is the most important factor in arena-elo?
Fit to the task beats raw benchmark rank. Define the failure modes, pick the benchmarks that resemble them, then test a small shortlist on your own examples.
How many models should I evaluate?
Three is usually enough: a quality candidate, a value candidate and a fallback. More candidates add work without changing the decision.
How do I control cost?
Compare blended 3:1 price, watch output tokens, use caching where available and route simple requests to a cheaper model.