Key takeaways
- skycrab1 set a random-seed glitchless world record of 6:39 on 27 July 2026, beating lowk3y_’s 6:50 from June 2025.
- It is the largest single improvement to the category record in over a year, and the first time the record has moved since the route was considered solved.
- Four of the five fastest runs in Minecraft history have now been set since the start of 2024.
The Minecraft world record moved on 27 July 2026. skycrab1 completed a random-seed glitchless run in 6:39, taking eleven seconds off lowk3y_’s 6:50 – a time that had stood since 11 June 2025 and had started to look like a wall rather than a record.
Eleven seconds is an enormous margin in a category where the entire run is under seven minutes and the top five are separated by less than half a minute. It is also the answer to the question the category had been asking for a year: whether the route had run out of room, or whether it just needed a seed generous enough to prove otherwise.
What the leaderboard looks like now
The top of the random-seed board has been rewritten from the top down. skycrab1’s 6:39 leads, lowk3y_’s 6:50 drops to second after thirteen months in front, and Jay12310’s 6:56 from February stays third.
What that table makes clear is how young the top of this leaderboard is. Four of the five fastest runs in the history of the game have been set since January 2024, in a category that has existed since 2019. The route did not get easier – the population of runners capable of executing it perfectly got much larger.
How a 6:39 is built
A record run in this category is a chain of cooperative dice rolls, each one of which can end the attempt. The seed has to put a village or ruined portal close to spawn for the opening kit. The nether portal has to be built from a lava pool rather than mined obsidian. The fortress has to be near the portal, the blazes have to drop rods early, and the bartering has to yield enough pearls in a single trip.
Each of those is a probability rather than a skill. What the runner contributes is the willingness to reset a few hundred worlds that fail any of them, and the execution to not lose seconds when all of them land at once.
Why the record stalled for thirteen months
Because the route stopped changing. The two structural improvements that defined the modern run – bastion-first pearl routing and the zero-cycle dragon kill – were both standard by the middle of 2025. Once every serious runner is executing the same optimal route, the difference between attempts is the world generation, not the plan.
That is why a stalled record is not evidence of a stagnant category. Through the whole of that stall, the times behind the record kept improving and the number of runners capable of a sub-seven run kept growing. The record was waiting on a seed, and enough attempts eventually produce one.
What eleven seconds actually buys
Broken down, eleven seconds in this category is roughly one of three things: a fortress that sits a few hundred blocks closer to the portal, a bartering trip that returns enough pearls on the first attempt instead of the second, or a stronghold that lands under the first eye throw rather than needing a correction. It is never all three, and it is almost never skill alone.
That is worth stating because record coverage tends to describe these runs as flawless. They are not – almost every record run in this category contains a small mistake the runner had enough margin to absorb. What separates a record from a good run is how much margin the seed handed over in the first place.
The ranked question
The obvious counterpoint: ranked completions have been coming in under six minutes for most of the year, and by that measure a 6:39 is not remarkable at all. It is a fair observation and a misleading one.
A ranked match is played on a shared seed the matchmaker generated for two players, and a fast ranked seed is a friendly one. A leaderboard run takes whatever the world generator produces and has to survive it. The numbers are not comparable, which is exactly why the categories are kept separate.
What happens next
Two things, probably. The first is a short flurry of near-misses: records tend to be followed by a cluster of runs within a few seconds, because the people chasing them raise their expectations at the same moment. The second is a long quiet, because the new bar is high enough that beating it will again require a seed most runners will never be handed.
The larger point is that a category widely described as solved just produced its biggest improvement in two years. Minecraft speedrunning has spent a decade being declared finished by people who then watched somebody take eleven seconds off it. Where the scene stood before this run is a useful reminder of how confident everyone was that 6:50 would hold.