> ## Content Index
> Fetch the complete content index at: https://aiadoption.org/llms.txt
> Use this file to discover other available public pages before exploring further.

# The best open-weight model trailed the best closed model by 49 Arena points in March 2026, down from 174
- URL: https://aiadoption.org/data/arena-closed-vs-open-weight-models/
- Published: 2026-10-19T15:55:52.000Z
- Updated: 2026-10-21T06:02:37.000Z
- Description: In March 2026 the top closed-weight model on the Arena scored 1,503 and the top open-weight model 1,454, a 49-point gap. In May 2023 the gap was 174 points; in February 2025 it was 5.
- Author: Manu
- Tags: Data, AI Models, Open Source AI, #chart, #ai-index-2026, #hub-models

1,454

▲ 27% since May 2023

Arena score of the top open-weight model in March 2026, 49 points behind the top closed-weight model (1,503)

20232026

## What the chart shows

The Arena ranks chatbots by users’ blind votes between two answers, with Elo-like scores. In March 2026 the top closed-weight model scored 1,503 and the top open-weight model 1,454: a gap of 49 points, or 3.4%. The report names them as Claude Opus 4.6 and GLM-5.

The gap has swung widely. It was 174 points in May 2023, narrowed to 7 in August 2024 and 5 in February 2025, then widened again as new closed models arrived, reaching 91 in May 2025.

## Key numbers

| Measure                           | Value |
| --------------------------------- | ----- |
| Top closed-weight model, Mar 2026 | 1,503 |
| Top open-weight model, Mar 2026   | 1,454 |
| Gap, Mar 2026                     | 49    |
| Gap, May 2023                     | 174   |
| Smallest gap, Feb 2025            | 5     |

## How to read it

- Each point is the highest score of any closed-weight or open-weight model that month on the Arena’s historical leaderboard (public, style control on), exported in March 2026.
- Arena scores reflect user preferences in blind comparisons, not accuracy on a fixed test. Model names come from the report text; the data gives scores only.
- Arena scores have no true zero, so the vertical axis starts at 1,100 to show the gaps; compare differences between lines, not their heights above the axis.

**Source:** [Arena](https://aiadoption.org/sources/#arena), 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.1.2: *Performance of top closed vs. open models on the Arena*. Chart and text by AI Adoption.  
Numbers checked against the published data on 20 October 2026.  
[Download the data (CSV)](https://charts.aiadoption.org/ai-index/csv/arena-closed-vs-open-weight-models.csv?ref=aiadoption.org) · [Cite this chart](#source)