SaaSBench (Engineering) - Data & Content Infrastructure: leaderboard

The Data & Content Infrastructure domain of SaaSBench (Ren et al., 2026): headless CMS, business intelligence, data catalog, feature-flag and web analytics systems, each built and deployed from scratch by a coding agent from a long product requirements document. Mean task score in percent over the domain tasks from the validation-graph checks, one rollout per task; higher is better.

Metric: Task score (%; pass@1, mean over the five Data & Content Infrastructure tasks of the share of each task's validation-DAG max score earned; agent builds and deploys an enterprise SaaS system from a PRD). Source: arxiv.org. Saturation forecast: Around August 2028. 16 models tracked.

Top models

#ModelScoreOverall rank
1Claude Opus 4.737.14#75
2GPT-5.4 (xHigh)14.5#121 (GPT-5.4)
3GLM-5.113.91#175
4MiniMax-M2.713.22#519
5DeepSeek V4 Pro12.78#181
6Kimi K2.610.51#173
7Gemini 3.1 Pro (Preview)9.97#91
8Qwen 3.6 Plus6.48#217

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=saasbench-engineering-data-content-infrastructure · How It Works · Data refreshed daily, snapshot 2026-11-12.