TSK-Bench · Build arena
Which AI model builds a working app from one request?
One request goes in, and the model must turn it into an app that opens, saves data and does what was asked.
Last test · 23 models · 55 builds
Results by request
Open a request to see every model ranked on it, with the request word for word.
| Request | Models | Works |
|---|---|---|
| Intake form Parents fill in a form on their phone, with photos, and a teacher sees every family on one page. Every required answer and every photo must save. | 23 | 12 |
| Game tracker app One Dota 2 player logs every match with the hero, the result and the stats. A dashboard shows win rate and streaks. | 14 | 9 |
No same-request table yet: Sales CRM, Inspection app, AI governance tracker, Reading list app, Habit tracker app, and Recipe app.
Every model across these requests
How many tested requests each model version got working. Each result comes from one to three builds, so read it as a direction.
| Model | Works on |
|---|---|
| 2 of 2 requests | |
| 2 of 2 requests | |
DeepSeek V4 Pro 0813 DeepSeek · Not in Taskade | 2 of 2 requests |
| 2 of 2 requests | |
| 2 of 2 requests | |
| 2 of 2 requests | |
| 2 of 2 requests | |
| 1 of 2 requests | |
| 1 of 2 requests | |
| 1 of 2 requests | |
| 1 of 2 requests | |
| 1 of 2 requests | |
| 1 of 1 request | |
| 1 of 1 request | |
GLM-5 Zhipu · Not in Taskade | 1 of 1 request |
| 1 of 1 request | |
| 1 of 1 request | |
| 0 of 2 requests | |
| 0 of 2 requests | |
Grok 4.20 xAI · Not in Taskade | 0 of 2 requests |
| 0 of 2 requests | |
DeepSeek V3.2 DeepSeek · Not in Taskade | 0 of 1 request |
GLM-4.7 Zhipu · Not in Taskade | 0 of 1 request |
| 0 of 1 request | |
Mistral Large 4 Mistral · Not in Taskade | 0 of 1 request |
More arenas coming: Automate (AI steps in workflows), Delegate (agents), Research (web research), Answer (chat), and Read (vision).
The full TSK-Bench ranking →Download the data (JSON)
How we grade these apps: read the full method.