Released in 2018, Celeste is a precision platformer developed and published by Maddy Makes Games, later operating under the Extremely OK Games, Ltd. name, and it launched across PC and major consoles, with later platform ports following.
As a speedgame, Celeste is built for repetition and refinement. The controls are crisp, the game is room-based with frequent checkpoints, and failures reset you instantly, which makes practice loops short and purposeful. Because outcomes are largely deterministic, runs are less about rolling good luck and more about sustained execution: carrying momentum through jumps, dashes, wall climbs, and corner interactions while keeping movement clean across dozens of tiny decision points.
What makes Celeste’s speedruns distinctive is how “micro” the optimization gets. Routes are usually framed around a clear chapter-to-chapter objective, but the real time save comes from threading fast, consistent movement through individual rooms and transitions. The community’s public record on speedrun.com reflects that structure with major full-game goals (like Any%, berry-focused goals, and other completion targets), and timing norms that center on the game’s built-in file or chapter time rather than real-time pacing.
Celeste’s speedrunning scene took shape almost immediately because the game shipped with a built-in “speedrun clock” that runners could enable without any external timing setup. Early guides aimed at first-time runners emphasized how simple it was to start: turn on the clock, choose File or Chapter timing, and run the story, with Any% framed in its most straightforward early form as “start a new file and play until you beat the Summit.”
From the beginning, record-keeping and discussion concentrated around the Celeste hub on Speedrun.com, which served as the public leaderboard and a living archive of runs, resources, and forum rulings. That same hub helped pull scattered learning efforts into shared tooling and documentation, including community-maintained splits and WR-progression tracking that made it easier to compare routes and eras once the first wave of records started stacking up.
Some of the earliest “standards” were less about movement tech and more about defining what, exactly, the timer meant. Community discussion quickly clarified that Celeste submissions were not meant to be ranked by real time at all, but by the in-game timers, with full-game runs using the file timer and IL runs using the level timer. Those expectations show up plainly in later Q and A threads: full-game submissions require the file timer (or an equivalent display via LiveSplit), and runners are told to leave real time blank and submit the chapter or file time shown by the game.
Even the Any% finish point was shaped by timing practicality. Early forum debate asked why Any% extended past Summit into Epilogue, and the answer was essentially “because the game’s timing display and file timer access make it the cleanest, most verifiable endpoint for everyone,” especially console runners. That same discussion also captured a foundational distinction the scene still leans on: Celeste effectively has two in-game timers, one used for individual chapters and one used for the full file, and the category norms grew around that reality.
As Celeste expanded and the community uncovered route-altering discoveries, those early standards evolved into clearer splits and tracking variables rather than constant reinvention. Speedrun.com news updates show the moderators responding to major Farewell-related glitches by creating sub-categories and clarifying how specific skips should be treated, while also using variables like version and dash-related flags to keep comparisons consistent across runs.
Celeste’s speedrunning community coalesced around speedrun.com as its public record. The game’s hub functions as more than a leaderboard. It is where category rules live, where moderators are listed as verifiers, and where forum threads accumulate the practical rulings that keep submissions consistent across platforms and versions. That structure matters for Celeste because so much of the competition is built on shared standards for timing, endpoints, and what constitutes a valid display of the in game clock.
As the scene grew, day to day discussion naturally spread beyond formal forums into faster social spaces. Community coordination increasingly happens through Discord, including large general Celeste hubs such as the Mt. Celeste Climbing Association server, with forum posts on the Speedrun.com page often pointing newcomers toward a current invite. At the same time, runners still use Reddit threads and Twitch chats as on ramps, places where beginners get directed to the same shared resources and communities that preserve route knowledge and answer rules questions.
Moderation and verification follow the familiar speedrun.com model: a runner submits a run through the game page’s submission form, typically with a video link as proof, and verifiers review the footage to confirm the category requirements and time are correct before accepting it to the leaderboard. For Celeste specifically, that process is tied to its timing identity. Submissions are expected to use the game’s chapter or file timer as the reported time, with real time left blank, which keeps comparisons consistent even when hardware and capture setups differ. Speedrun.com’s own support guidance also sets expectations for the human pace of volunteer moderation, including a recommended waiting period before pinging moderators about an unreviewed submission.
Knowledge in the Celeste scene is preserved in a practical, teachable way. The speedrun.com hub collects guides and tools that standardize how people train and present runs, including setup instructions for things like autosplitters, while community discussion channels reinforce the same shared vocabulary for routes, settings, and timing displays. The result is a culture where information is meant to be repeatable. A new runner can start with basic submission norms, then steadily layer in the tooling and technique that turn “beating the game fast” into a disciplined craft.
On speedrun.com, Celeste’s full game leaderboards are built around a familiar set of “goal” categories, with Any%, All Red Berries, True Ending, All Cassettes, Bny%, All Hearts, and 100% as the main spine, plus longer completion variants like 202 Berries and All Chapters listed as miscellaneous categories. The site also supports individual level leaderboards alongside the full game boards, which matters because Celeste’s timing conventions depend on what you are running.
What gets measured is IGT, and the community leans on the game’s own timers rather than pure real time. Celeste complicates this in an interesting way because it effectively has two in-game timing contexts: the on-screen chapter timer behavior that is relevant to level runs, and the file timer used for full game categories. In community discussion, runners point out that the overall file timer is the one used for full game runs and that it only fully “stops” in a way you can verify after the Epilogue, which is why full game categories traditionally include the Epilogue as the practical endpoint for timing and submission.
The biggest rule decisions tend to show up where new skips or strategies threaten comparability. A clear example is how the community chose to handle Farewell Skip in full game categories by splitting True Ending and All Chapters into subcategories, with the primary board disallowing the skip and a secondary board allowing it. For certain golden berry style categories, the approach is not always a full split, but a variable noting whether the skip was used. Even presentation requirements can become “rules” when they affect verification, like the note that heavy pause buffering plus a journal bind can hide the in-game timer, in which case a visible timer with proper decimals is required so the run can still be verified.
Platform and version matter, too, because submissions are tracked with both fields on the leaderboard tables, which helps the community keep runs comparable and transparent when releases differ. And while the main boards aim to represent standard, unassisted play, Celeste also has an organized outlet for intentional rule-bending as its own “separate lane,” including Category Extensions boards for things like Assist Mode and Cheat Mode, rather than mixing those runs into the core categories.
Celeste speedrunning matured in a way that matches the game’s design. Because each chapter is built from discrete rooms with clear entry and exit points, early routing quickly shifted from “which path is shortest” to “how do I flow through every room with as few hesitations as possible.” The long term gains came from shaving micro-pauses, reducing climbs and stamina waste, and turning awkward setups into repeatable lines that could survive tournament nerves and marathon pressure.
That push for consistency is what made movement tech so central. Over time, runners leaned harder on a family of advanced dashes and speed-carrying options such as supers, hypers, wavedashes, and ultras, plus more situational interactions like cornerboosts and wallbounce variants, all of which let you keep momentum through room transitions and turn recovery moments into forward motion. Community documentation tends to describe this as a timing and spacing discipline rather than a single trick, with attention to when dash cooldowns end, when a jump can restore a dash, and how specific surfaces and objects convert velocity into new routes.
As the knowledge base expanded, the scene also professionalized its practice habits. The standard toolkit for serious runners includes modded practice environments, most famously the Speedrun Tool, which supports save states, room timers, and detailed death and time-loss feedback so runners can grind the same room until it becomes automatic. Those tools commonly ride on the Everest mod loader ecosystem, which helped make “lab work” a normal part of Celeste training rather than something only a few runners could set up.
On the presentation and pacing side, Celeste runs benefitted from the same timing quality-of-life improvements that many modern games did. The community maintains autosplitter setup guidance and curated split packs so runners can track chapters, checkpoints, or IL structure cleanly without manual bookkeeping, which reinforces a culture of analysis and repeatability instead of purely gut-feel pacing.
Finally, the biggest “paradigm shifts” tend to arrive when a discovery is so powerful it threatens to redefine what the run even is. Celeste has seen that with major Farewell-related skips and dash-retention glitches that forced the community to clarify how full game and IL boards should treat them, sometimes by creating subcategories so different philosophies can coexist without erasing each other’s history.
One of Celeste’s earliest landmark milestones was the creation of a shared, public record that felt native to the game itself. The built in speedrun clock gave runners a common language for performance, and speedrun.com quickly became the place where categories, submissions, and standards could stabilize in public rather than being scattered across personal spreadsheets and one off videos. That combination did more than track times. It set expectations for what a run should look like, how it should be timed, and how it should be presented across different platforms and versions.
As the scene matured, “barrier breaking” became a recurring milestone type that signaled new confidence in both route and execution. Celeste has had clear sub X moments in major categories that functioned as shared reference points, not because a number is magic on its own, but because crossing it usually meant a route had become coherent enough, and consistent enough, that multiple runners could reproduce elite pacing rather than one person surviving a perfect attempt. Runs and discussions around the first sub 30 era for Any% captured that shift in a way the community still recognizes as a turning point in what was considered realistically repeatable play.
Marathon showcases became another kind of milestone, where the community’s private knowledge base had to be translated into an audience friendly story. Celeste’s appearances at Awesome Games Done Quick and Summer Games Done Quick helped cement a public facing version of the run, one that emphasized readable routing, clear explanations of movement, and the rhythm of room based execution. Those stages also gave the game a wider cultural footprint inside speedrunning, turning Celeste from a popular run into a recurring reference point for precision movement games.
A different kind of watershed moment came when new content and new mechanics forced the scene to renegotiate what “standard” meant. The release of Chapter 9, Farewell, introduced ideas and interactions that quickly became central to high level play, and it also created pressure on category identity as powerful skips and route bending strategies emerged. The community response, including formal rule updates and category splits around Farewell Skip in certain full game boards, became a milestone in its own right because it showed the moderators prioritizing historical continuity and fair comparison rather than letting one discovery overwrite every earlier era.
Finally, Celeste’s most enduring milestones are often the tools and documents that made its knowledge portable. Community maintained technique references, practice tooling, and mod support turned improvement into something you could study systematically instead of rediscovering through trial and error. When a runner can isolate a single room, drill a specific movement family, and measure the result with consistent timing and splits, that is more than convenience. It is how the scene preserves hard won understanding and makes the next generation of route overhauls possible.
secureaccount – Any% – 24 m 51 s 495 ms – 1 year ago – PC / 1.4.0.0 – A modern top-board benchmark for the post-routing, movement-first era.
Valyne – Any% – 25 m 03 s 820 ms – 7 months ago – PC / 1.4.0.0 – A clean example of how “small things everywhere” still stack into major full-run pace.
Isaactayy (Sweden) – Any% – 26 m 04 s 884 ms – Date not shown in this capture – PC / 1.4.0.0 – A recognizable “26 era” run that sits close enough to elite pace to illustrate how execution, not route, is the limiter.
Zkad – Any% – 26 m 14 s 523 ms – 4 years ago – PC / 1.4.0.0 – Often cited in community retellings as part of the grind that made sub-26 feel inevitable rather than mythical.
TGH – Any% – 29 m 59 s 178 ms – 7 years ago – PC / 1.2.4.1 – A foundational barrier run that defined an early “sub-30 is real” moment for the category.
Isaactayy (Sweden) – Any% – 25 m 59 s 070 ms – 3 years ago – Platform/version not listed on this page – The first widely celebrated “first 25” milestone, treated as a watershed achievement by the broader speedrun community.
Isaactayy (Sweden) – All Red Berries – 44 m 47 s 972 ms – 2 years ago – PC / 1.4.0.0 – A flagship completion-style record where route knowledge and long-run consistency matter as much as raw movement.
yujene – 100% – 1 h 38 m 09 s 004 ms – 10 months ago – PC / 1.4.0.0 (DTS: Yes) – A modern standard for the long-form “do everything fast” side of the scene, shaped heavily by endurance execution.
yujene – All Chapters (No Farewell Skip) – 1 h 22 m 45 s 564 ms – 10 months ago – PC / 1.4.0.0 (DTS: Yes) – A defining full-game showcase run under the “no Farewell skip” policy, reflecting how the community wants the main board to represent the game.
Isaactayy (Sweden) – True Ending (No Farewell Skip) – 43 m 52 s 671 ms – 10 months ago – PC / 1.4.0.0 (DTS: Yes) – A marquee mid-length category that highlights clean A/B-side execution while preserving the intended “ending” boundary.
GrosHiken – Farewell (Clear, Farewell Skip) – 22 m 57 s 162 ms – 3 months ago – PC / 1.4.0.0 (DTS: Yes) – A modern IL benchmark for the game’s most demanding chapter, and a good “what elite Celeste looks like” reference point.
Rickfernello – 202 Berries – 4 h 55 m 11 s 467 ms – 3 years ago – PC / 1.4.0.0 – A completion marathon that represents the extreme edge of the community’s endurance categories.
Isaactayy (Sweden) – All Cassettes – 33 m 59 s 235 ms – 4 years ago – PC / 1.4.0.0 – A historically notable “collectathon” pace-setter, and a reminder that subgoals create their own optimization ecosystems.
Maddy Makes Games. “Celeste – Changelog – v1.4.1.0.” Celeste (official site). Accessed February 2, 2026. https://www.celestegame.com/changelog.html
Maddy Makes Games. “v1.4.0.0 Changelog.” itch.io (devlog). Accessed February 2, 2026. https://maddymakesgamesinc.itch.io/celeste/devlog/237097/v1400-changelog
Steam. “Celeste – V1.4.0.0 Changelog.” Steam News. Accessed February 2, 2026. https://store.steampowered.com/news/app/504230/view/2984171597628988822
Steam. “Celeste.” Steam Store page. Accessed February 2, 2026. https://store.steampowered.com/app/504230/Celeste/
Speedrun.com. “Celeste.” Game hub and leaderboards. Accessed February 2, 2026. https://www.speedrun.com/celeste
Speedrun.com. “Celeste Leaderboards: Any%.” Accessed February 2, 2026. https://www.speedrun.com/celeste?h=Any&x=7kjpl1gk
Speedrun.com. “Celeste – Levels.” Individual level leaderboards. Accessed February 2, 2026. https://www.speedrun.com/celeste/levels
Speedrun.com. “Celeste – Resources.” Tools, splits, and community resources. Accessed February 2, 2026. https://www.speedrun.com/celeste/resources
Speedrun.com. “How to set up the autosplitter for Celeste.” Guide. Accessed February 2, 2026. https://www.speedrun.com/celeste/guides/nmhul
Speedrun.com. “File / Level Timer.” Forum thread on full-game timing expectations. Accessed February 2, 2026. https://www.speedrun.com/celeste/forums/v9zdv
Speedrun.com. “How to you count the ‘real time’ version of the timer?” Forum thread clarifying IGT-only submission norms. Accessed February 2, 2026. https://www.speedrun.com/celeste/forums/cbcx3
Speedrun.com. “Ending the run on Summit instead of Epilogue in Any%.” Forum discussion of practical endpoints and timer visibility. Accessed February 2, 2026. https://www.speedrun.com/celeste/forums/m3cul#rlojm
Speedrun.com. “Handling of Farewell, Rules Update, and New Modded Board.” News post documenting major category/rules decisions. Accessed February 2, 2026. https://www.speedrun.com/celeste/news/nx9ql439
Speedrun.com. “Leaderboard Changes due to Farewell Glitch.” News post documenting DTS-era leaderboard structure. Accessed February 2, 2026. https://www.speedrun.com/celeste/news/kyxy953n
celeste.ink. “Tech.” Celeste community wiki technique reference. Accessed February 2, 2026. https://celeste.ink/wiki/Tech
Everest. “Everest – Celeste Mod Loader.” Installation and modding entry point. Accessed February 2, 2026. https://everestapi.github.io/
GitHub. DemoJameson. “Celeste.SpeedrunTool.” Repository for the Speedrun Tool practice mod. Accessed February 2, 2026. https://github.com/DemoJameson/Celeste.SpeedrunTool
Reddit. speedyskier22. “Guide to speedrunning Celeste for those who have never speedrun (or played, really).” r/celestegame. Accessed February 2, 2026. https://www.reddit.com/r/celestegame/comments/a7tln2/guide_to_speedrunning_celeste_for_those_who_have/
Reddit. “First 25 Celeste Any% in 25:59.070 by isaactayy.” r/speedrun. Accessed February 2, 2026. https://www.reddit.com/r/speedrun/comments/yzobfi/first_25_celeste_any_in_2559070_by_isaactayy/
Games Done Quick. “Celeste.” Game page with curated marathon VOD links. Accessed February 2, 2026. https://gdqvods.com/game/celeste
YouTube. Games Done Quick. “Celeste by Lmjacks and yujene in 1:01:15 – Summer Games Done Quick 2025.” Video. Accessed February 2, 2026. https://www.youtube.com/watch?v=dMn6YJ-0fPs