GameASG-Bench: leaderboard
Metric: Strict task success rate (%; tasks out of all 47 planned browser-game generation tasks whose delivered index.html passes every L1 source-level check and every applicable L2 prerequisite and core-requirement behavioral check in headless Chromium; one run per task from a clean workspace with full tool access; read from the printed success count as count x 100/47). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6-Astra (Codex CLI, max) | 55.32 |
| 2 | Claude-Opus-5 (Claude Code, max) | 51.06 |
| 3 | GPT-5.6-Sol (Codex CLI, max) | 44.68 |
| 4 | DeepSeek-V4-Flash (Claude Code, high) | 40.43 |
| 5 | DeepSeek-V4-Flash (Claude Code, max) | 38.3 |
| 6 | DeepSeek-V4-Flash (Codex CLI, max) | 38.3 |
| 7 | DeepSeek-V4-Pro (Claude Code, max) | 31.91 |
| 8 | Kimi-K3 (Claude Code, max) | 31.91 |
| 9 | GLM-5.2 (Claude Code, max) | 23.4 |
| 10 | Hunyuan-3 (Claude Code, max) | 21.28 |
Interactive version: theaggregate.ai/benchmark?slug=gameasg-bench · How It Works · Data refreshed daily, snapshot 2026-09-26.