Hello,
in 2024 we reported to QGIS issues with abusive levels of OpenStreetMap tile usage from some QGIS users causing QGIS's tile usage to be excessive. In October 2024 QGIS implemented changes that they believed would mitigate the issues.
In May 2025 we re-opened the issue on our end as we were seeing the same problems but did not have the time to do a detailed investigation at that point.
Recent times when we have hit capacity limits have lead us to look at QGIS usage in more detail. For this message I looked at data from four weekdays1: 2026-03-04, 2026-03-05, 2026-03-06, and 2026-03-09. QGIS' usage of OSM tile resources now significantly exceeds OpenStreetMap's usage of our own resources.
QGIS has been averaging 1,794 tiles per second (TPS) with a miss rate of 473 TPS. Cache misses are more expensive to serve and what puts load on our backend servers. Prohibited behavior like scraping or downloading tiles for offline usage shows up as a high cache miss rate.
To put QGIS' usage into context, I pulled data for it and five other large categories of users
| User | TPS | TPS CDN miss | Miss rate |
|---|---|---|---|
| Mozilla/5.0 QGIS/* | 1,794.0 | 473.0 | 26.4% |
| openstreetmap.org | 1,779.5 | 297.1 | 16.7% |
| flutter_map (*) | 1,608.0 | 67.0 | 4.2% |
| Chrome 145 with no referer | 557.3 | 38.1 | 6.8% |
| localhost | 588.2 | 40.3 | 6.9% |
| openrailwaymap.org | 284.5 | 28.5 | 10.0% |
QGIS' usage now exceeds that of osm.org itself and has become excessive. Its high miss rate indicates a noticable portion of its traffic is from users scraping2.
It appears that the 2024 changes were added in 3.40. They did not help. Of the 1726 TPS of z13+ tiles, 1112 TPS were from 3.40 or later.3
Looking specifically into if it was a small number of QGIS users I looked at IPs downloading over 50,000 z13+ tiles in the same hour at some point over the four days. I found 25 osm.org users4 and 44 QGIS users. This is better than previous, likely because we have implemented more rate limits.
We are looking at increased server-side rate-limiting that will impact QGIS users and reduce the problems. Unfortunately, all the options are blunt tools that will impact the experience for all QGIS users.
Ultimately, QGIS is responsible for the traffic from QGIS. It is essential that the QGIS project takes steps to identify the ways QGIS users are breaking the standard tile layer terms of service and take measures to inhibit them.
Please let us know if there is any additional information we can provide that will help you identify the causes of problems.
Footnotes
-
QGIS usage is significantly higher on weekdays, more so than other traffic. ↩
-
A typical app or site has a miss rate of 1% to 8%. A site where the user is zoomed in and looking at recent edits will be 10% or so. openstreetmap.org is higher at 16.7% because it uses different caching logic. QGIS' 26.4% miss rate indicates some users are scraping with it. ↩
-
z13+ tiles are not pre-rendered and scraping them causes more load problems than other zooms. They are also more likely to be scraped in the first place. ↩
-
There is evidence about half of the claimed osm.org users above the use threshold were fakes and not from osm.org. We have no evidence of systemtatic fake QGIS users. ↩