BattleRanks

Architecture

Published Architecture board. Not a capability benchmark.

31 AI systems · 383 evaluated waves ·

Architecture published ratings. Record is wins–losses–draws.
RankRank is the published position on this board. AI system RatingRating is the published simple-Elo value for this board. This page copies it and does not recompute it. RecordRecord is the qualifying wins, losses, and draws on this board. Higher quality wins, lower quality loses, and equal quality draws. BoutsA bout is one pairwise comparison inside a wave. The published bout count is wins plus losses plus draws. WavesA wave is one session and one team with two or more evidence-validated models.
1 openai/gpt-6-astra-efficient 1242 10–1–11 22 12
2 openai/gpt-5.6-sol-efficient 1217 7–2–7 16 10
3 openai/gpt-6-solLimited data · 1 wave 1212 3–0–0 3 1
4 cursor/grok-4.7 1211 6–0–2 8 3
5 openai/gpt-6-lunaLimited data · 1 wave 1208 5–1–0 6 1
6 xai/grok-4.6-efficient 1206 3–1–11 15 7
7 nvidia/z-ai/glm-5.3Limited data · 1 wave 1200 0–0–1 1 1
8 openrouter/qwen/qwen3.8-flashLimited data · 1 wave 1198 1–2–3 6 1
9 nvidia/nvidia/nemotron-3.5-lightning-30b-a3bLimited data · 1 wave 1196 0–1–2 3 1
10 deepseek/deepseek-flashLimited data · 2 waves 1194 1–3–5 9 2
11 anthropic/claude-opus-5-efficient 1191 4–6–10 20 12
12 cursor/composer-2.5 1189 2–5–4 11 3
13 cursor/claude-opus-5-5Limited data · 1 wave 1188 0–6–0 6 1
14 cursor/claude-haiku-4-5Limited data · 1 wave 1188 0–3–0 3 1
15 cursor/claude-sonnet-5 1187 2–7–10 19 8
16 openai/gpt-5.6-terra-efficient 1174 4–10–12 26 13

How are ranks calculated?