Vision2Web S4-2026 (Codex) - Level 3 Full-Stack: leaderboard

Vision2Web Level 3: the agent develops a full-stack web application from requirements and prototypes; the score is the mean of a VLM-judged visual score and a functional score from workflow-driven GUI tests, on a 0-100 scale under the Codex framework, higher is better (He et al., 2026). Season S4-2026 (Oct to Dec 2026) uses Gemini-3.5-flash as the visual judge and GLM-5V-Turbo (Claude Code) as the GUI testing agent; scores compare only within a season.

Metric: Score (0-100). Source: vision2web-bench.github.io. Saturation forecast: Around February 2027. 3 models tracked. Updated 4 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.6 Sol60.6#36
2GPT-5.6 Terra56.2#72
3GPT-5.6 Luna52.6#133

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

Interactive version: theaggregate.ai/benchmark?slug=vision2web-s4-2026-codex-level-3-full-stack · How It Works · Data refreshed daily, snapshot 2026-11-20.