InteractWeb-Bench: leaderboard

InteractWeb-Bench (2026) tests website-generation agents that may ask a simulated user, played with one of four personas, to clarify ambiguous, redundant or contradictory requirements before building a site. This board is the task completion rate in percent over easy, middle and hard tasks; higher is better.

Metric: Task completion rate TCR (%) of interactive website generation, where the agent may ask a simulated user with one of four personas for clarification before building the site, over easy, middle and hard tasks; higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.6 Plus38.78#238
2Kimi K2.538.26#266
3Qwen 3.5 397B A17B37.15#264
4Gemma 4 31B (IT)36.96#350
5GPT-4.134.33#505
6Gemma 4 26B A4B (IT)30.39#556
7GPT-4.1 Mini29.39#789
8Gemini 3.1 Flash Lite25.72#371
9Qwen 3.5 9B24.33#828

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=interactweb-bench · How It Works · Data refreshed daily, snapshot 2026-11-21.