|
Limit |
| Product URL |
2,048 characters, http or https, no username or password in it |
| Focus |
500 characters |
| Reading the page |
One page, up to 2 MB, 20 seconds, 5 redirects |
| Text given to the setup model |
The first 12,000 characters |
| What setup proposes |
Up to 4 people and 6 goals |
| New projects per workspace |
10 in 10 minutes, and 100 in 24 hours, failed attempts included |
|
Limit |
| People |
1 to 12 |
| A person’s name |
100 characters |
| What they are like |
2,000 characters |
| Goals |
1 to 20 |
| A goal |
1,000 characters |
| Test accounts |
Up to 20 per project |
| A test account’s username |
320 characters |
| A test account’s password |
8 to 1,000 characters |
|
Limit |
| Hard cap |
$0.10 to $50, in steps of $0.10; $2 by default |
| Model key |
20 to 400 characters of letters, digits, ., _ and - |
| Model name |
200 characters |
| Listing models |
30 times in 10 minutes per person; up to 1,000 models listed |
| The key check |
One call of at most 5 tokens, 30 seconds |
|
Limit |
| A person’s session |
60 steps |
| A replay |
30 steps |
| Sessions at a time |
One per run: people take their turn one after another |
| A model call |
Up to 16,000 tokens of output, 180 seconds |
| A model with no known price |
3 million tokens, input and output together |
| A session that stops reporting |
Given up after 10 minutes |
| A finding’s title |
300 characters |
| What a person saw |
4,000 characters |
| A finding’s steps |
1 to 30, at least 2 for a defect |
| A note a person writes |
4,000 characters |
The local runner has its own defaults — 120 steps per person, 40 per replay and a $5 budget — which its options change.