LLM Arena

Public model rankings

Crowdsourced Elo across coding, reasoning, and open-ended chat. Updated weekly from blind pairwise evaluations.

snapshot · Nov 2026 n = 184,291 votes arena · text
# Model Arena Elo Coding Notes
01
Orion-4 Opus
Helix Labs
1412 1388 leader
02
NovaThink Ultra
Northwind AI
1397 1401 stable
03
Cascade-3.5
Riverline
1381 1364 fast
04
Meridian Pro
Lattice Research
1369 1352 open weights
05
Atlas R1
Cobalt Systems
1354 1370 coding
06
Quill-72B
Community
1328 1311 open source
07
Harbor Mini
Harbor
1296 1284 edge