Best for coding
Claude Opus 5 leads · 2 fresh boards · counts toward the leader of leaders · Fixing real repositories, building web apps, editing code.
12 of 12 models
| # | Model | Score#1s | ||
|---|---|---|---|---|
| 1 |  Claude Opus 5Anthropictop 10 on 2, #1 on 1 | 85.0 | 85.01 | 2 |
| 2 | Claude Fable 5.1Anthropictop 10 on 2 | 85.0 | 85.00 | 2 |
| 2 |  GPT-6 AstraOpenAItop 10 on 2 | 85.0 | 85.00 | 2 |
| 4 |  Kimi K3Moonshot AItop 10 on 2 | 55.0 | 55.00 | 2 |
| 5 | Claude Opus 5.5Anthropictop 10 on 1, #1 on 1 | 50.0 | 50.01 | 1 |
| 6 | Claude Sonnet 5Anthropictop 10 on 1 | 35.0 | 35.00 | 1 |
| 7 | GPT-6 SolOpenAItop 10 on 1 | 30.0 | 30.00 | 1 |
| 8 |  Qwen3.8 Max (0902)Alibabatop 10 on 1 | 25.0 | 25.00 | 1 |
| 9 | GPT-5.6 TerraOpenAItop 10 on 1 | 25.0 | 25.00 | 1 |
| 10 |  GLM-5.3Z.ai (Zhipu)top 10 on 1 | 20.0 | 20.00 | 1 |
| 11 |  Muse Spark 1.3Metatop 10 on 1 | 15.0 | 15.00 | 1 |
| 12 | GPT-5.6 SolOpenAItop 10 on 1 | 15.0 | 15.00 | 1 |
Boards
| Board | Score | ||
|---|---|---|---|
 WebDev ArenaLMArena#1 Claude Opus 5.5 | Claude Opus 5.5 | 1,818 | Sep 23, 2026 |
 SWE-Bench ProScale AI#1 Claude Opus 5 | Claude Opus 5 | 98.0% | Sep 24, 2026 |
 SWE-bench VerifiedSWE-bench#1 Claude Opus 4.5 | Claude Opus 4.5 | 76.8%stale | Feb 26, 2026 |
 Aider PolyglotAider#1 GPT-5 | GPT-5 | 88.0%stale | Oct 3, 2025 |
The wall
#1 with the vendor’s logo23 podium7 top 10 stale board, not countedArrow keys move, Enter opens
Rank over time
Gemini 3 ProScores as published by each board on the capture date. Model names and logos belong to their owners; logos via logo.dev.