Design Arena: agents building games leaderboard
- Claude Fable 5 leads
- 1285
- models measured
- 26
- lead over #2
- +19 Elo
- median
- 1180
Leader on Design Arena: agents building games · read October 22
On October 22, 2026, 26 of our catalog's models had a Design Arena: agents building games result, and Anthropic Claude Fable 5 led with 1285, ahead of OpenAI GPT-5.6 Sol at 1266.
Design Arena shows people two designs made by two models from the same prompt (a game, built by the model as an agent) and asks which is better. The number is the Elo rating built from those votes, as Design Arena publishes it.
Role in the index
Design Arena: agents building games is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Battles | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5Anthropic | 1285 | 1,428 | 79.3 | $50.00 | |
| 2 | GPT-5.6 SolOpenAI | 1266 | 1,054 | 73.5 | $20.00 | |
| 3 | Claude Opus 5Anthropic | 1261 | 789 | 83.0 | $25.00 | |
| 4 | Kimi K3Moonshot AI | 1252 | 700 | 62.3 | $15.00 | |
| 5 | Claude Opus 4.8Anthropic | 1239 | 1,382 | 65.0 | $25.00 | |
| 6 | Grok 4.5xAI | 1236 | Not published | 58.6 | $6.00 | |
| 7 | Gemini 3.7 FlashGoogle | 1235 | 4,716 | 58.0 | $3.75 | |
| 8 | Claude Sonnet 5Anthropic | 1227 | 1,417 | 57.7 | $10.00 | |
| 9 | Claude Opus 4.6Anthropic | 1212 | 224 | 57.2 | $25.00 | |
| 10 | Grok 4.6xAI | 1209 | Not published | 65.7 | $6.00 | |
| 11 | GLM 5.2Z.ai | 1206 | 855 | 61.1 | $4.40 | |
| 12 | Muse Spark 1.1Meta | 1183 | Not published | 70.3 | $4.25 | |
| 13 | Gemini 3.5 FlashGoogle | 1180 | 3,588 | 39.8 | $9.00 | |
| 14 | Gemini 3.6 FlashGoogle | 1180 | 1,700 | 62.5 | $3.75 | |
| 15 | Muse Spark 1.2Meta | 1179 | 624 | 55.7 | $4.25 | |
| 16 | GPT-5.5OpenAI | 1179 | 2,275 | 71.3 | $30.00 | |
| 17 | Claude Sonnet 4.6Anthropic | 1177 | 245 | 57.3 | $15.00 | |
| 18 | GLM 5.1Z.ai | 1164 | 1,499 | 49.2 | $4.40 | |
| 19 | Qwen3.7 MaxQwen | 1161 | 714 | 54.1 | $4.425 | |
| 20 | MiniMax M3MiniMax | 1158 | 527 | 39.2 | $1.20 | |
| 21 | Kimi K2.6Moonshot AI | 1141 | 2,462 | 52.1 | $4.00 | |
| 22 | Kimi K2.7 CodeMoonshot AI | 1130 | 832 | 53.4 | $4.00 | |
| 23 | Gemini 3.1 Pro PreviewGoogle | 1113 | 1,870 | 32.3 | $12.00 | |
| 24 | GLM 5V TurboZ.ai | 1102 | 1,953 | Not benchmarked | $4.00 | |
| 25 | GPT-5.6 TerraOpenAI | 1083 | 668 | 63.9 | $12.00 | |
| 26 | Grok 4.3xAI | 1007 | 1,921 | 37.4 | $2.50 |
Only models in our catalog are listed, and each appears once with its best published run. 145 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.