GTA-Workflow - Root Score: leaderboard
Metric: Mean root checkpoint score (0-10) 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 December 2027. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 3.66 |
| 2 | Llama 4 Scout | 3.65 |
| 3 | Gemini 2.5 Pro | 3.64 |
| 4 | Qwen 3 235B A22B | 3.59 |
| 5 | DeepSeek V3.2 | 3.56 |
| 6 | Grok 4 | 3.56 |
| 7 | Claude Sonnet 4.5 | 3.5 |
| 8 | Kimi K2 | 3.5 |
| 9 | Qwen 3 8B | 1.81 |
| 10 | Llama 3.1 70B Instruct | 1.55 |
| 11 | Qwen 3 30B A3B | 1.21 |
| 12 | Llama 3.1 8B Instruct | 1.18 |
| 13 | Llama 3.2 3B Instruct | 1.02 |
Interactive version: theaggregate.ai/benchmark?slug=gta-workflow-root-score · How It Works · Data refreshed daily, snapshot 2026-10-07.