UniClawBench - Multimodal: leaderboard

Metric: Pass rate (%) on the 80 multimodal tasks of 400 bilingual real-world tasks run live in Docker containers under the OpenClaw framework (v2026.3.11), with up to two simulated user follow-ups and a hidden GPT-5.4 Codex supervisor that grades step checkpoints. Source: arxiv.org. Saturation forecast: Around June 2028. 10 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.621.2
2GPT-5.417.5
3Claude Opus 4.817.5
4Gemini 3.1 Pro (Preview)16.3
5Gemini 3 Flash8.7
6GPT-4.17.5
7Kimi K2.67.5
8Qwen 3.5 Plus7.5
9GPT-5.4 Mini6.2
10Gemini 3.1 Flash Lite3.7

Interactive version: theaggregate.ai/benchmark?slug=uniclawbench-multimodal · How It Works · Data refreshed daily, snapshot 2026-09-29.