Zcash NU7 block times, expected and counted

When does NU7 reach Zcash mainnet?

Block 3,543,405 is fixed in Zebra 7.0.0. On 28 October 2026, 11:28 UTC our header client stood at block 3,533,510, with 9,895 blocks to go. At the pace we counted over the 14 full UTC days up to 27 October 2026, 75.29 s per block, the median of 20,000 draws puts block 3,543,405 at 6 November 2026, 02:26 UTC. That is our estimate.

Nine in ten draws fall between 5 November 2026, 23:01 UTC and 6 November 2026, 05:51 UTC. At exactly 75 s per block it would be 6 November 2026, 01:37 UTC. All three are estimates computed from our count, not dates set by the project.

Our own header client, Zcash mainnet, as of 28 October 2026. Method in How we measure.

Measurement uncertainty. The 90% band is the spread of the draws from the pace we counted. It does not cover a change in hashrate after 28 October 2026, so we publish each new forecast with its date and score it on our track record.

Methods used PH-M-001, PH-M-006.

What is the Zcash block time before NU7?

Before NU7 the mainnet targets 75 s per block. From 20 September 2026 to 28 October 2026 we counted 43,793 intervals, with a mean of 75.34 s and a median of 53.0 s, close to an ideal 75 s chain.

Our own header client, Zcash mainnet, blocks 3,489,718 to 3,533,510, as of 28 October 2026.

Measurement uncertainty. The interval column is the 95% range of each value, from the rank statistics for percentiles and the standard error for the mean. Header times are whole seconds set by the miners.

Methods used PH-M-001.

Does NU7 hit 25 second blocks on testnet?

Yes. The mean interval was 23.68 s, so blocks came about 5% faster than the target. The median of 17.0 s is what a 25 s chain should show, because most blocks arrive early and a few arrive late.

Computed means an ideal chain with a 25 s mean, where intervals follow an exponential distribution.

Our own measurement, Zcash NU7 testnet, headers from testnet peers, blocks 4,465,026 to 4,514,228, snapshot 18 October 2026, 06:00 UTC.

Measurement uncertainty. The interval column is the 95% range of each value, from rank statistics, the standard error or a Wilson interval. Header times are whole seconds set by the miners.

Methods used PH-M-001.

How often is a block slower than the old 75 s target?

In 3.59% of intervals, against 4.98% for an ideal 25 s chain. Before NU7 the target itself was 75 s. Gaps over 150 s came in 0.203% of intervals, against 0.248% expected.

The longest gap was 451 s, longer than the roughly 284 s an ideal chain would show over the same number of blocks. No block timestamp went backwards.

Our own measurement, Zcash NU7 testnet, headers from testnet peers, blocks 4,465,026 to 4,514,228, snapshot 18 October 2026, 06:00 UTC.

Measurement uncertainty. The 95% intervals for both shares are in the table above. The longest gap is a single counted value.

Methods used PH-M-001.

How fast does difficulty follow hashrate?

We found 6 hashrate jumps of 2x or more, and 5 of them caught up. The other one was followed by the next jump 200 blocks later, before it could catch up. For those five the difficulty came within ±15 % of the new level after a median of 960 blocks, or about 2.7 hours when measured in time, and the longest catch-up took 1,839 blocks.

Our own measurement, Zcash NU7 testnet, headers from testnet peers, blocks 4,465,026 to 4,514,228, snapshot 18 October 2026, 06:00 UTC.

Measurement uncertainty. Five catch-ups are too few for an interval, so the median of 960 blocks is a single counted value. Difficulty is read from the nBits field of each header and is exact.

Methods used PH-M-001.

How long do confirmations take?

Thirty blocks at 25 s spanned a median of 730 s and a p99 of 1,187 s. For perfect 25 s blocks the p99 would be about 1,105 s, so the measured value sits above that, partly because of the hashrate jumps (1,154 s without the windows around them), and still well below the 1,409 s for ten blocks at 75 s.

Our own measurement, Zcash NU7 testnet, headers from testnet peers, blocks 4,465,026 to 4,514,228, snapshot 18 October 2026, 06:00 UTC.

Measurement uncertainty. Spans of 30 blocks overlap and are not independent, so they get no sample interval. Each span carries the whole second resolution of two header times.

Methods used PH-M-001.

How many blocks were orphaned?

From 17 October 2026, 14:51 to 20:21 UTC our listener recorded every block announced to it. Five of the 780 did not land in the chain, 0.64%, and each of them at depth 1.

Our own listener, Zcash NU7 testnet, window 17 October 2026, 14:51 to 20:21 UTC.

Measurement uncertainty. The 95% Wilson interval for 5 of 780 is 0.27 to 1.49%. A block our listener never heard of is not counted, so the share is a lower bound for what the whole network saw.

Methods used PH-M-002, PH-M-005.

How we measured

Our own header client followed the Zcash testnet from the activation of NU7 (ZIP 259) at block 4,465,026 to block 4,514,228, the snapshot of 18 October 2026, 06:00 UTC, fetching headers directly from testnet peers and checking hash linkage and proof of work. That is 49,203 blocks. We checked 500 of them against two Zebra peers, and all 500 hashes matched, so we saw no fork in the blocks we compared.

Intervals come from consecutive block timestamps. Method in How we measure, version 1.

Download the raw intervals

The file comes from our own header client, which fetched every block header directly from testnet peers over P2P, checked the hash linkage and the proof of work against nBits, and compared 500 hashes with two further Zebra peers. It does not check block contents or Equihash solutions, and it does not come from nodes of our own.

• height is the block height on the Zcash testnet, from 4,465,026 to 4,514,228.
• timestamp is the block header timestamp in Unix seconds, which count in UTC.
• interval_s is that timestamp minus the timestamp of the parent block, in seconds. It stays empty for 4,465,026, because the first counted interval is that of 4,465,027.

Download CSV, 49,203 blocks, about 1 MB

Data from PulseHawk, licensed CC BY 4.0. Cite as PulseHawk (2026), Zcash NU7 testnet block intervals, cut-off 18 October 2026.

Our own measurement, Zcash NU7 testnet, headers from testnet peers, blocks 4,465,026 to 4,514,228, snapshot 18 October 2026, 06:00 UTC. SHA-256 c1cda3d79a470f5a8e508ebb8ba633dd4f3499eec79b5ecbba96b446dea603b6.