Providers
Compare RPC providers on measured data
The top 4 on the chosen chain are compared by default. Pick your own — up to 4 — and the selection moves into the URL, so a comparison can be sent to someone else and they will see exactly what you saw.
Providers on Ethereum
Side by side
30d window · 2.3 days observed · generated 2026-08-27 04:46:15 UTC
| Metric | NodeReal | Merkle | Cloudflare |
|---|---|---|---|
| Rank | 4 | 6 | 9 |
| Score | 88.6 | 78.9 | 55.6 |
| Status | Good | Good | Degraded |
| Availability | 99.80% | 99.98% | 23.10% |
| p50 latency | 87.3 ms | 106.5 ms | 15.1 ms |
| p95 latency | 99.8 ms | 195.9 ms | 88.3 ms |
| p99 latency | 123.1 ms | 244.4 ms | 253.7 ms |
| Correctness | 99.99% | 100.00% | 100.00% |
| Head lag | 0.35 blocks | 0.47 blocks | no cohort |
| Rate limited | 0 | 14,629 | 77 |
| Unsupported | 0 | 180 | 2,671 |
| Samples | 89,367 | 21,039 | 90,720 |
9 providers scored against each other.
Where the gap comes from
Two scores side by side say one is higher. This says why: the score is a weighted sum, so it comes apart, and each component’s share of the difference adds back to the difference exactly. Compared against NodeReal, the highest scorer of those selected.
NodeRealvsMerkle
+9.7points
- Latency+9.6
- Head freshness+0.1
- Availability−0.1
- Correctness−0.0
The parts add to +9.7, exactly — the score is a weighted sum, so it comes apart.
NodeRealvsCloudflare
+33.0points
- Availability+29.6
- Head freshness*+9.6
- Latency−4.0
- Correctness−2.2
The parts add to +33.0, exactly — the score is a weighted sum, so it comes apart. Components marked * were measured for only one of the two; the other’s weight is spread across what it does have, which is why the two sides can carry different weights for the same row.