
We suspect most of the improvements noted on the prior page, such as with Vodafone, will be due to customers upgrading to full fibre (FTTP) based packages from slower ADSL and FTTC (SOGEA / VDSL2) lines – quite a few of the major providers still have plenty of customers on legacy connections left to convert. But quite why EE and BT didn’t seem to convert this into better overall performance remains unclear. Even Zen suffered a drop, albeit mostly in upstream.
Next year may well bring even faster FTTP products from Openreach, although it’s not yet known whether those will become commercially available by the end of 2026 or how many ISPs will adopt them. We are also expecting Virgin Media to launch an even faster tier than 2Gbps at some point (5Gbps has long been hinted), which may also impact future results.
The following table reflects a top list of the fastest smaller providers (excluding the major providers listed above), which also includes some large-ish providers that may not have strong national availability due to the limited coverage of their underlying network(s). This is intended to provide some useful context against the biggest and most widely available market players above.
Advertisement
The table below is ordered by median download speed, but for extra context we’ve also included a figure for upload performance in brackets. Most of the providers in this list are faster than the major players because they’re often connected to alternative networks with faster entry-level speeds and will have a low portion of users (sometimes none at all) on older copper-based lines. But there are some exceptions (e.g. some highly rated ISPs that have been around a long time may suffer in this table due to the impact of legacy users on slower lines).
Top 20 Smaller ISPs – Average Downloads (Upload)
| Operator | H2 – 2025 | H1 – 2025 |
| toob | 447.6Mbps (404.4Mbps) | 508Mbps (379.2Mbps) |
| Aquiss | 438.7Mbps (94.9Mbps) | 908.7Mbps (108.3Mbps) |
| CommunityFibre | 427.2Mbps (302.3Mbps) | 506.9Mbps (373.9Mbps) |
| brsk | 389.8Mbps (354.4Mbps) | 239.2Mbps (252.8Mbps) |
| YouFibre | 375.6Mbps (302.4Mbps) | 287.2Mbps (197.5Mbps) |
| Hyperoptic | 360.3Mbps (269.3Mbps) | 394Mbps (355Mbps) |
| Trooli | 332.1Mbps (187.7Mbps) | no data (new entry) |
| Hey! Broadband | 308.2Mbps (437.5Mbps) | 404.5Mbps (401Mbps) |
| Connect Fibre | 272.6Mbps (252.2Mbps) | 260.7Mbps (230.2Mbps) |
| Highland Broadband | 233.1Mbps (127.9Mbps) | 230.2Mbps (225Mbps) |
| Squirrel Internet | 225.4Mbps (125Mbps) | 346.8Mbps (262.9Mbps) |
| Wessex Internet | 222.4Mbps (130.1Mbps) | no data (new entry) |
| Gigaclear | 204.8Mbps (198.5Mbps) | 236.8Mbps (209.9Mbps) |
| Ogi | 200.9Mbps (30.4Mbps) | no data (new entry) |
| Zzoomm | 192.9Mbps (156.2Mbps) | no data (new entry) |
| Fibrus | 171.9Mbps (85Mbps) | 237.2Mbps (77Mbps) |
| Quickline | 150.4Mbps (115.7Mbps) | no data (new entry) |
| Andrews and Arnold | 147.8Mbps (37Mbps) | no data (new entry) |
| iDNET | 130Mbps (73.5Mbps) | no data (new entry) |
| KCOM | 106.3Mbps (105.2Mbps) | 242.1Mbps (104.9Mbps) |
Overall, the average speeds shown above are significantly faster than many – but not all – of the market’s largest ISPs. This isn’t surprising as a lot of the smaller players will be using modern alternative full fibre networks that have managed to pull customers away from often slower packages and older connection technologies at the biggest providers.
The catch is that smaller providers produce a smaller data sample, which means that these results will fluctuate more over time. Internet providers are also more likely to both dip into and out of the above list, between updates, due to many of them only just being able to produce the minimum required level of test data.
Advertisement
For example, in this update we had to remove LightSpeed Broadband, Cuckoo, Lightning Fibre, LitFibre, Grain Connect, WightFibre, County Broadband and 4th Utility due to lack of data. But in their place we added Trooli, Wessex Internet, Ogi, Zzoomm, Quickline, Andrews & Arnold (AAISP), iDNET and KCOM.
ISPreview has also been keeping track of the results for SpaceX’s satellite based Starlink broadband service. Sadly, not enough data exists to include other satellite platforms or providers, but that may change in the future. Otherwise, Starlink’s average download performance appears to have risen strongly during the second half of 2025 and it’s a similar story for their upload performance.
| H2 – 2025 (Top 10%) | H1 – 2025 (Top 10%) | |
| Download | 126.2Mbps (199.4Mbps) | 78.7Mbps (223.4Mbps) |
| Upload | 19.2Mbps (26.2Mbps) | 10.8Mbps (18.5Mbps) |
Mobile speeds remain a difficult thing to study because end-users are always moving through different areas (indoor, outdoor and underground), using different devices with different capabilities and the surrounding environment is ever changeable (weather, trees, buildings etc.). All of this can impact signal quality and that’s before we consider any differences in network (backhaul) capacity or spectrum usage between locations.
Suffice to say that studies of mobile broadband speed are inherently open to variation, but the top networks often tend to be those with a combination of the best 4G or 5G coverage, a good amount of radio spectrum and the most advanced technologies.
Advertisement
Average Mobile Download Speeds – H2 2025 vs H1 2025
| No. | Operator | H2 2025 (Top 10%) | H1 2025 (Top 10%) | Change |
| 1. | Three UK | 34.9Mbps (275.3Mbps) | 36.1Mbps (268.1Mbps) | -3.32% |
| 3. | EE | 26.8Mbps (126.8Mbps) | 31.5Mbps (124.3Mbps) | -14.92% |
| 2. | Vodafone | 25.8Mbps (149.6Mbps) | 20.6Mbps (98.7Mbps) | 25.24% |
| 4. | O2 | 19.5Mbps (118.9Mbps) | 17.5Mbps (51.2Mbps) | 11.43% |
Average Mobile Upload Speeds – H2 2025 vs H1 2025
| No. | Operator | H2 2025 (Top 10%) | H1 2025 (Top 10%) | Change |
| 1. | EE | 6.4Mbps (33.3Mbps) | 7.6Mbps (35Mbps) | -15.79% |
| 2. | Three UK | 5.4Mbps (34.2Mbps) | 6.2Mbps (41.6Mbps) | -12.9% |
| 3. | Vodafone | 3.9Mbps (13.1Mbps) | 4.2Mbps (18.3Mbps) | -7.14% |
| 4. | O2 | 3.1Mbps (12.5Mbps) | 3Mbps (17.3Mbps) | 3.33% |
Overall, the average download speed of the four primary mobile operators was 26.75Mbps (up from 26.42Mbps in H1 2025) and the average upload speed hit 4.7Mbps (down from 5.25Mbps). Both EE and Three UK seemed to suffer a notable performance decline, with Vodafone also losing out a bit on uploads, although variations like this aren’t uncommon for mobile operators.
The market for 5G services in the UK is now starting to reach a degree of maturity, with swings of c.5-10% in one direction or the other becoming somewhat common between biannual updates. But at the same time, we’re a little be surprised to see that the newer generation of faster Standalone 5G (SA) networks don’t seem to be having much of an overall performance impact, yet.
Going forward, it will be interesting to see how the gradual merger of Vodafone and Three UK’s respective networks impacts these results. At present we’re still listing them both as separate operators, but in the next year or two we may start to combine their results into a single score, although more network integration will need to happen first.
Finally, we’re not expecting Ofcom’s recent auction of 5G spectrum in the mmW bands – 26GHz and 40GHz – to have much of an impact. The mobile operators have yet to introduce related services and there’s a significant lack of device support. On top of that, those makes really only make sense for busy urban areas or events, thus they probably won’t move the overall performance dial by much.
Privacy Notice: Please note that news comments are anonymous, which means that we do NOT require you to enter any real personal details to post a message and display names can be almost anything you like (provided they do not contain offensive language or impersonate a real person's legal name). By clicking to submit a post you agree to storing your entries for comment content, display name, IP and email in our database, for as long as the post remains live.
Only the submitted name and comment will be displayed in public, while the rest will be kept private (we will never share this outside of ISPreview, regardless of whether the data is real or fake). This comment system uses submitted IP, email and website address data to spot abuse and spammers. All data is transferred via an encrypted (https secure) session.