GTA-Workflow - Perception: leaderboard

Root success on the GTA-Workflow tasks that require perception tools, part of GTA-Workflow, the open-ended tier of GTA-2 (2026): 132 long-horizon productivity workflows from real user queries with deployed tools and multimodal inputs, run end to end in a ReAct agent framework and judged by GPT-5.2 against recursive checkpoint rubrics. Percent; higher is better.

Metric: Root success rate (%) on the tasks that require perception tools on GTA-Workflow, the open-ended long-horizon workflow tier of GTA-2 (132 real-user tasks with deployed tools and multimodal inputs), run end to end in the default Lagent ReAct framework; deliverables judged by GPT-5.2 against recursive checkpoint rubrics; higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 235B A22B15.79#628
2Llama 4 Scout15.79#1818
3Gemini 2.5 Pro13.16#276
4GPT-513.16#178
5Claude Sonnet 4.510.53#261
6DeepSeek V3.210.53#419
7Kimi K210.53#443
8Grok 47.89#319
9Llama 3.1 70B Instruct2.63#1571
10Qwen 3 30B A3B2.63#1156
11Llama 3.1 8B Instruct0#4600
12Qwen 3 8B0#1873
13Llama 3.2 3B Instruct0#7241

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

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