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

#ModelScore
1GPT-6-Astra (Codex CLI, max)55.32
2Claude-Opus-5 (Claude Code, max)51.06
3GPT-5.6-Sol (Codex CLI, max)44.68
4DeepSeek-V4-Flash (Claude Code, high)40.43
5DeepSeek-V4-Flash (Claude Code, max)38.3
6DeepSeek-V4-Flash (Codex CLI, max)38.3
7DeepSeek-V4-Pro (Claude Code, max)31.91
8Kimi-K3 (Claude Code, max)31.91
9GLM-5.2 (Claude Code, max)23.4
10Hunyuan-3 (Claude Code, max)21.28

Interactive version: theaggregate.ai/benchmark?slug=gameasg-bench · How It Works · Data refreshed daily, snapshot 2026-09-26.