Revisions to The Complete Rate-Limiting Guide

added 1 character in body

Source Link

edited Jun 15 at 2:13

96.8k
17
193
413

API request quotas
- Request quotas are the maximum number of API requests in a rolling 24-hour period.
- Maximum of 300 queries per 24 hours per IP address, if not using an API key value. Requests with an API key also count against this quota. It is not separate from the 10k quota, which means these 300 requests are likely to be consumed by applications using an API key.
- Maximum of 10,000 queries per day per API key and IP address pair, if using a per-application API key, with some exceptions:
  - Some high-profile tools that actively help out with the Stack Exchange network receive special keys that grant them an extended limit (e.g., SmokeDetector has a limit of 80100,000)
  - The above noted exemption for the iOS app, which has a special key that doesn't have daily API limits (see above section for more details)
  - The quota which is applied to the IP address is the same for all API requests made with a key value in the 24-hour period. The quota value which is used is the maximum quota assigned to any key which is used to make a request from that IP address within that 24 hour period (e.g., if a key with a 50k quota is used for a single request on the IP address, then all requests with a valid key value from that IP address in that 24 hour period have a 50k quota, regardless of what their normal quota would be).
- Requests which use an access_token, in addition to a key, use a quota which is tied to the access_token. An access_token uniquely identifies an application + user pair. Each user has up to 5 additional 10,000 request quota pools which are allocated separately to the first 5 application + user pairs used for that user in a 24-hour period. Requests made with an access_token do not count against the IP based quotas.
A hard limit of 30 requests from any IP address per second. SE considers "> 30 request/sec per IP to be very abusive and thus cut the requests off very harshly." In normal operation, requests will almost always be throttled in other ways well before that hard limit.
Occasionally, queries can return a backoff value, to indicate how much time a given application needs to wait before it can perform another query. This value depends on the number and speed of the queries and the overall load on SE's servers, and can change based on various other factors. The exact criteria used for issuing a backoff are not disclosed.
SE uses multiple layers of rate limiting for the SE API. The criteria used for the rate limiting layers are not disclosed. Rate limiting is implemented both by the above mentioned backoff response and requests returning various errors, potentially with no additional information.

API request quotas
- Request quotas are the maximum number of API requests in a rolling 24-hour period.
- Maximum of 300 queries per 24 hours per IP address, if not using an API key value. Requests with an API key also count against this quota. It is not separate from the 10k quota, which means these 300 requests are likely to be consumed by applications using an API key.
- Maximum of 10,000 queries per day per API key and IP address pair, if using a per-application API key, with some exceptions:
  - Some high-profile tools that actively help out with the Stack Exchange network receive special keys that grant them an extended limit (e.g., SmokeDetector has a limit of 80,000)
  - The above noted exemption for the iOS app, which has a special key that doesn't have daily API limits (see above section for more details)
  - The quota which is applied to the IP address is the same for all API requests made with a key value in the 24-hour period. The quota value which is used is the maximum quota assigned to any key which is used to make a request from that IP address within that 24 hour period (e.g., if a key with a 50k quota is used for a single request on the IP address, then all requests with a valid key value from that IP address in that 24 hour period have a 50k quota, regardless of what their normal quota would be).
- Requests which use an access_token, in addition to a key, use a quota which is tied to the access_token. An access_token uniquely identifies an application + user pair. Each user has up to 5 additional 10,000 request quota pools which are allocated separately to the first 5 application + user pairs used for that user in a 24-hour period. Requests made with an access_token do not count against the IP based quotas.
A hard limit of 30 requests from any IP address per second. SE considers "> 30 request/sec per IP to be very abusive and thus cut the requests off very harshly." In normal operation, requests will almost always be throttled in other ways well before that hard limit.
Occasionally, queries can return a backoff value, to indicate how much time a given application needs to wait before it can perform another query. This value depends on the number and speed of the queries and the overall load on SE's servers, and can change based on various other factors. The exact criteria used for issuing a backoff are not disclosed.
SE uses multiple layers of rate limiting for the SE API. The criteria used for the rate limiting layers are not disclosed. Rate limiting is implemented both by the above mentioned backoff response and requests returning various errors, potentially with no additional information.

API request quotas
- Request quotas are the maximum number of API requests in a rolling 24-hour period.
- Maximum of 300 queries per 24 hours per IP address, if not using an API key value. Requests with an API key also count against this quota. It is not separate from the 10k quota, which means these 300 requests are likely to be consumed by applications using an API key.
- Maximum of 10,000 queries per day per API key and IP address pair, if using a per-application API key, with some exceptions:
  - Some high-profile tools that actively help out with the Stack Exchange network receive special keys that grant them an extended limit (e.g., SmokeDetector has a limit of 100,000)
  - The above noted exemption for the iOS app, which has a special key that doesn't have daily API limits (see above section for more details)
  - The quota which is applied to the IP address is the same for all API requests made with a key value in the 24-hour period. The quota value which is used is the maximum quota assigned to any key which is used to make a request from that IP address within that 24 hour period (e.g., if a key with a 50k quota is used for a single request on the IP address, then all requests with a valid key value from that IP address in that 24 hour period have a 50k quota, regardless of what their normal quota would be).
- Requests which use an access_token, in addition to a key, use a quota which is tied to the access_token. An access_token uniquely identifies an application + user pair. Each user has up to 5 additional 10,000 request quota pools which are allocated separately to the first 5 application + user pairs used for that user in a 24-hour period. Requests made with an access_token do not count against the IP based quotas.
A hard limit of 30 requests from any IP address per second. SE considers "> 30 request/sec per IP to be very abusive and thus cut the requests off very harshly." In normal operation, requests will almost always be throttled in other ways well before that hard limit.
Occasionally, queries can return a backoff value, to indicate how much time a given application needs to wait before it can perform another query. This value depends on the number and speed of the queries and the overall load on SE's servers, and can change based on various other factors. The exact criteria used for issuing a backoff are not disclosed.
SE uses multiple layers of rate limiting for the SE API. The criteria used for the rate limiting layers are not disclosed. Rate limiting is implemented both by the above mentioned backoff response and requests returning various errors, potentially with no additional information.

80 reviews per First Questions queue per day if the queue size is 1000 or more on Stack Overflow // https://meta.stackoverflow.com/a/412077/839601

Source Link

edited Jun 28, 2023 at 22:09

gnat

11.2k
26
149
328

20 reviews per queue per day if the queue size is less than 1000 (150 on Stack Overflow)
40 reviews per queue per day if the queue size is 1000 or more (150 on Stack Overflow)
80 reviews per First Questions queue per day if the queue size is 1000 or more on Stack Overflow

Reviews on suggested edits to your own posts do count toward the limit if you haven't reached it yet, but will be allowed through even if you have.
No review limit for ♦ mods

Rollback to Revision 148 - Rollback of prior edit. The edit uses a singular example to denote the 30minute rate limit of SO was reversed and back to 3, but uses an example that actually shows answers made less than 3 minutes apart too. https://meta.stackexchange.com/a/385820/397219

Source Link

edited Jan 14, 2023 at 23:26

Larnu

1k
6
16

Users with < 125 rep must wait 3 minutes between answers on most Stack Exchange sites, but must wait 30 minutes between answers on Stack Overflow. The longer wait time between answers on Stack Overflow was implemented as part of the response to a flood of ChatGPT answers.
Users with between 125 and 10k rep trip CAPTCHA* if more than once per 60 seconds, or within 5 seconds of starting new post
Users with ≥ 10k rep trip CAPTCHA* if more than once per 30 seconds, or within 5 seconds of starting new post
Users of any reputation level can only answer the same question once every 60 seconds
Users with the "staff" bit (i.e. Stack Exchange employees) are not subject to the above limits on meta sites

Seems the rate limit for Stack Overflow has returned to normal.

Source Link

edited Jan 14, 2023 at 22:11

Robert Longson

35.7k
13
82
163

Loading

added 171 characters in body

Source Link

edited Jan 12, 2023 at 9:25

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

deleted 36 characters in body

Source Link

edited Jan 10, 2023 at 23:36

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

Delay between answering on SO increased to 30 minutes (2022-12-08).

Source Link

edited Dec 30, 2022 at 19:37

Makyen

24.5k
8
43
79

Loading

More detail regarding the SE API quotas and actual rate limits.

Source Link

edited Jun 22, 2022 at 15:27

Makyen

24.5k
8
43
79

Loading

SD's limit was extended

Source Link

edited Jun 22, 2022 at 8:36

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

managed to run into this today

Source Link

edited Jun 22, 2022 at 3:35

bobble

8.3k
4
21
62

Loading

See prior edit summary

Source Link

edited Jun 6, 2022 at 6:29

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

Rollback to Revision 137 - This isn't the correct place to put character limits; this is only for rate limits (actions that can only be taken at a certain frequency). /questions/372184/an-incomplete-character-limiting-guide-to-stack-exchange is the right place for that, and all of these additions are mentioned there.

Source Link

edited Jun 6, 2022 at 6:22

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

added 110 characters in body

Source Link

edited Jun 6, 2022 at 5:53

Rebecca J. Stones

22.3k
5
45
104

Loading

added 129 characters in body

Source Link

edited Jun 6, 2022 at 5:41

Rebecca J. Stones

22.3k
5
45
104

Loading

added 129 characters in body

Source Link

edited Jun 6, 2022 at 5:33

Rebecca J. Stones

22.3k
5
45
104

Loading

This is for deletion, too.

Source Link

edited May 25, 2022 at 6:56

Ryan M

28.1k
9
77
129

Loading

https://meta.stackexchange.com/a/371277/1017231 - "This change won't affect the ability of moderators to delete/undelete a post multiple times, if necessary."

Source Link

edited Oct 26, 2021 at 19:18

bobble

8.3k
4
21
62

Loading

added 36 characters in body

Source Link

edited Oct 26, 2021 at 19:09

Luuklag

36.1k
7
73
176

Loading

added 72 characters in body

Source Link

edited Oct 26, 2021 at 18:59

Sonic the Anonymous Hedgehog

96.8k
17
193
413

Loading

Rollback to Revision 131 - This is already covered in the FAQ, in the one on how reputation works. This isn't really a *rate* limit, or a limit as to how many actions a user can take at a time.

Source Link