WeaveBench - Overall Score: leaderboard

Metric: Mean per-task score (0-100) on 114 long-horizon real-world computer-use tasks that need GUI, CLI, code and API tools together, each model run as the backbone of a fixed OpenClaw agent runtime with the full hybrid tool pool at its best thinking mode; each deliverable is graded by a five-layer rubric pipeline whose correctness tier caps the task score, printed 0-1 and x100 here; higher is better. Source: arxiv.org. Saturation forecast: Around August 2027. 10 models tracked.

Top models

#ModelScore
1Claude Opus 4.7 (High)48.2
2GPT-5.5 (High)46.6
3GPT-5.4 (High)46.5
4GPT-5.3 Codex (High)45.6
5Qwen 3.5 397B A17B31.8
6GPT-5.1 Codex (High)22.6
7Gemini 3.1 Pro (Preview)22.3
8Qwen 3 VL 8B (Thinking)9.2

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