GTA-Workflow - Root Score: leaderboard

Mean root checkpoint score on 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. Score from 0 to 10; higher is better.

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 June 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-53.66#178
2Llama 4 Scout3.65#1818
3Gemini 2.5 Pro3.64#276
4Qwen 3 235B A22B3.59#628
5DeepSeek V3.23.56#419
6Grok 43.56#319
7Claude Sonnet 4.53.5#261
8Kimi K23.5#443
9Qwen 3 8B1.81#1873
10Llama 3.1 70B Instruct1.55#1571
11Qwen 3 30B A3B1.21#1156
12Llama 3.1 8B Instruct1.18#4600
13Llama 3.2 3B Instruct1.02#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-root-score · How It Works · Data refreshed daily, snapshot 2026-11-21.