GTA-Workflow - Logic: leaderboard
Metric: Root success rate (%) on the tasks that require logic 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 June 2028. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 13.93 |
| 2 | GPT-5 | 12.3 |
| 3 | Llama 4 Scout | 11.48 |
| 4 | Grok 4 | 10.66 |
| 5 | Qwen 3 235B A22B | 10.66 |
| 6 | Claude Sonnet 4.5 | 9.84 |
| 7 | DeepSeek V3.2 | 9.84 |
| 8 | Kimi K2 | 8.2 |
| 9 | Llama 3.1 70B Instruct | 0.82 |
| 10 | Llama 3.2 3B Instruct | 0.82 |
| 11 | Qwen 3 30B A3B | 0.82 |
| 12 | Llama 3.1 8B Instruct | 0 |
| 13 | Qwen 3 8B | 0 |
Interactive version: theaggregate.ai/benchmark?slug=gta-workflow-logic · How It Works · Data refreshed daily, snapshot 2026-10-07.