Chat2Workflow: leaderboard
Metric: Resolve rate (%) on Chat2Workflow (79 instructions in 27 multi-turn tasks, three test cases each, pooled over six domains): share of test cases where the workflow the model generates for the current round of a multi-turn request, deployed on Dify 1.9.2, executes and its output satisfies every requirement of the instruction, judged by GPT-5.1; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 60.2 |
| 2 | Claude Sonnet 4.5 | 47.96 |
| 3 | GPT-5.2 | 47.4 |
| 4 | GLM-4.7 | 46.55 |
| 5 | Kimi K2 (Thinking) | 36 |
| 6 | GLM-4.6 | 35.02 |
| 7 | GPT-5.1 | 34.46 |
| 8 | DeepSeek V3.1 | 32.07 |
| 9 | DeepSeek V3.2 | 27.43 |
| 10 | Kimi K2 | 24.61 |
| 11 | Qwen 3 235B A22B | 24.47 |
| 12 | Qwen 3 32B | 19.13 |
| 13 | Qwen 3 Coder 480B A35B Instruct | 19.13 |
| 14 | Qwen 3 14B | 12.38 |
| 15 | Qwen 3 8B | 6.61 |
Interactive version: theaggregate.ai/benchmark?slug=chat2workflow · How It Works · Data refreshed daily, snapshot 2026-10-07.