Arena coding leaderboard

Arena coding standings (134 models)

Sorted by published score, one row per model. Source: Arena (LMArena).

How ModelCap uses Arena coding

Arena coding ranks models by human preference rating on coding prompts. ModelCap ingests the board as published, matches each entry to a catalogue model with a reviewed identity, and shows the source's score, interval and rank unchanged. Where a model has an admitted result, it feeds the ModelCap Index as coding evidence alongside the other public boards; the methodology documents the weighting and the identity rules.

Other benchmark leaderboards

• Arena text leaderboardClaude Fable 5 leads 114 models at 1506
• LMArena Agent leaderboardClaude Opus 5 leads 34 models at 12.0%
• ARC-AGI-2 leaderboardGPT-5.6 Sol leads 49 models at 92.5%
• ARC-AGI-3 leaderboardClaude Opus 5 leads 9 models at 30.2%
• SWE-bench leaderboardGemini 3 Flash Preview leads 13 models at 75.8%
• BFCL V4 function-calling leaderboardGLM 4.6 leads 18 models at 72.4

Explore further

• AI model rankingsThe live ModelCap Index across every board
• All benchmark leaderboards7 public boards
• MethodologyHow boards become the ModelCap Index
• Claude Fable 5 vs Kimi K3Head-to-head on Arena coding, price and context
• Claude Fable 5 vs Muse Spark 1.2Head-to-head on Arena coding, price and context
• Claude Fable 5 vs Claude Opus 5Head-to-head on Arena coding, price and context
• Kimi K3 vs Muse Spark 1.2Head-to-head on Arena coding, price and context
• Anthropic modelsPublisher of the Arena coding leader