AgentRoboCritic - Action Outcome Accuracy: leaderboard

AgentRoboCritic (arXiv:2611.26613v1, Table 1): zero-shot multimodal action-level critic accuracy (%) on predicting action outcome across 400 human-annotated direct robot-control actions (off-the-shelf models only); higher is better.

Metric: Action Outcome Accuracy (%). Source: arxiv.org. Saturation forecast: Around March 2027. 10 models tracked. Updated 21 Nov 2026.

Top models

#ModelScoreOverall rank
1GPT-6 Astra79.25#10
2Claude Fable 5.278.25#1
3GPT-6.1 Sol74.5#15
4GPT-6.1 Terra74.5#25
5Claude Opus 5.572.5#9
6GLM-5.3 Flash62#133
7Claude Sonnet 5.556.75#24
8Kimi K3.1 Pro54.5#36
9DeepSeek V4.1 Flash (1017)54.5#47
10Muse Glimmer 30B52.5#398

No result here: #3 Gemini 4 Neon, #4 GPT-6.2 Astra, #7 Gemini 4 Carbon.

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