Minecraft 26.3 changes how the game draws anything you can see through. Stained glass, water, ice, leaves, and any block layered behind another translucent block now go through a different rendering pipeline than they did in 1.21.4 or 26.2. The feature is called Order-Independent Transparency, OIT for short, and it quietly replaces a setting most players never touched: Improved Transparency.

This page covers what OIT is as a technique, what it fixes, what it costs, and what it means for shader packs, all pinned to Pre-Release 1 (1 September 2026), the build where 26.3's feature set froze. Anything below marked as verified comes from that build or from the Snapshot 2 changelog that introduced OIT; anything we could not verify is labeled as such rather than guessed at.

RateVerified against Pre-Release 1
XP levelJava 26.3
Build timeChecked: 1 September 2026
DifficultyIntermediate

Why this is settled, not a snapshot rumor. OIT landed in Snapshot 2 (30 June 2026) and survived unchanged through eight more snapshots into Pre-Release 1. Pre-releases are bug-fix only: Mojang is not adding or cutting features at this stage, only polishing what already shipped. Pre-Release 1 did touch OIT once, adding an alternative depth-copy path for hardware the original method didn't support, covered below. That is a compatibility fix, not evidence the feature is still moving.

What Improved Transparency actually did wrong

To understand what changed, it helps to know what broke under the old system.

The classic GPU pipeline draws opaque objects first, then translucent ones on top, back to front, so the color behind a translucent surface shows through correctly. That's easy for one pane of glass. It falls apart once you stack several translucent surfaces the GPU can't cheaply sort per pixel: two sheets of angled stained glass, water behind glass, leaves over ice. The renderer guesses an order, and when it guesses wrong you get the symptom every player has actually seen: a translucent block flickering between showing and hiding what's behind it as the camera moves, or one pane in a stack rendering as opaque, or missing, depending on the angle.

Improved Transparency was Mojang's earlier fix, using extra framebuffers to composite layers more carefully (the same idea behind the old "Fabulous" graphics setting). It helped, but the root problem, sorting translucent geometry by draw order, was still there. OIT fixes the actual root cause: it's a rendering technique that composites translucent fragments without sorting them by draw order at all. Instead of hoping the GPU draws layers in the right sequence, it accumulates every translucent fragment landing on a pixel and combines them mathematically so the result is correct regardless of arrival order. That's the "order-independent" part of the name, and it's why Mojang's own Snapshot 2 changelog calls OIT "expected to be much better at solving translucency issues" than what it replaced.

Improved Transparency (old) OIT (26.3)
How it composites layers Extra framebuffers, still draw-order dependent Accumulates fragments per-pixel, no draw-order dependency
Stacked glass/water flicker Present, worsens with more layers Targeted directly by the algorithm
Where it lives in the renderer Its own post-processing pass New core/oit_composite.fsh and supporting oit_*.glsl shader files
Toggle in Video Settings Yes, "Improved Transparency" On/Off No dedicated settings-menu toggle found; a debug keybind (F3 + X) exists instead

Is Improved Transparency still a setting you can turn off?

Here's what we could verify and what we couldn't. The Snapshot 2 changelog is explicit that OIT "replaced" the Improved Transparency approach, not that it sits beside it as an alternative. The same changelog documents a new debug keybind, F3 + X by default, for toggling OIT on and off. A debug keybind isn't a Video Settings option: debug binds exist for developers and testers to compare behavior, not as a persistent player-facing setting saved to your options file.

We did not find a wiki entry or changelog line stating outright that the "Improved Transparency" checkbox was removed from the Video Settings menu. What we can say with confidence: the rendering approach it used to control has been replaced outright, and the only toggle documented in the 26.3 changelogs is the F3 + X debug bind, not a menu option. If you're on 26.3, check Video Settings for an "Improved Transparency" entry yourself; if it's gone, OIT is unconditional. We'll update this page once we can confirm against a release build.

This is the one open question on this page. Everything else here is sourced directly to the Pre-Release 1 changelog or the Snapshot 2 changelog that introduced OIT. This specific point, whether a menu toggle survives, is our best reading of the evidence rather than a confirmed fact.

The Pre-Release 1 detail that matters for older or unusual hardware

Pre-Release 1 added a new shader file, core/blit_depth.fsh, described in the changelog as an alternative method for copying the contents of depth textures, added specifically because the existing rendering-backend-based copying approach "does not work on some devices."

That's a small technical line with a real implication. Depth-texture copying is plumbing OIT depends on to know what's in front of what before compositing translucent layers. Shipping a second implementation of that plumbing this late in the cycle, purely for compatibility, means the original method had devices it didn't work correctly on. We don't know which devices, how many players it affected, or whether it caused a crash, a visual glitch, or something subtler, because Mojang's changelog doesn't say. But it's a concrete signal that OIT's rollout wasn't friction-free on every GPU and driver combination, worth knowing if you're on older or less common hardware, especially older integrated graphics or less-maintained drivers. If translucent blocks look wrong specifically around glass, water, or leaves after updating, this is the first thing to suspect, and updating your graphics drivers is the first thing to try.

Does OIT cost FPS?

Here's the honest answer: no public benchmark comparing OIT to the old Improved Transparency pipeline exists yet, and this page will not invent one. Pre-Release 1 was one day old as of this writing, and nobody has published FPS numbers for it that we could verify.

What we can say qualitatively, and this is reasoning, not a measured result: order-independent transparency techniques generally work by accumulating multiple fragments per pixel instead of writing one, which means more memory bandwidth and more per-pixel work than a simple sorted blend. That's the general tradeoff this class of technique makes across game engines, not something specific we measured in Minecraft. Whether it shows up as a real frame-rate hit depends on how much translucent geometry is on screen: a little water and a few glass panes should barely register, while a base built mostly out of stained glass, or a view looking through several stacked layers at once, is exactly where a heavier compositing pass would cost the most. That mirrors why Fabulous graphics, which made a similar "extra work to get transparency right" tradeoff, was always the most GPU-expensive setting in the game.

If you're on a weaker or older GPU and 26.3 feels slower specifically in transparency-heavy views rather than everywhere, OIT is a reasonable first suspect. The troubleshooting playbook doesn't change: check whether you're CPU-bound or GPU-bound with F3 and a hardware overlay, and if it's GPU-side, the complete Minecraft performance guide covers the settings that matter. Until someone publishes a proper before-and-after benchmark, treat any specific FPS percentage you see elsewhere as unverified.

What it means for shader packs

This is the question the update is most likely to break for a specific, sizable group of players, and it's also where we have the least to report.

Iris has no published 26.3 build as of 2 September 2026, verified against Iris's own Modrinth version page. Its latest release still targets 26.2. Nobody, including us, can currently run a shader pack against Minecraft 26.3, because the shader loader itself doesn't run on it yet.

Here's why this particular change is worth watching once Iris ships a 26.3 build. Shaders hook deep into the rendering pipeline, and transparency compositing is exactly the kind of pass a pack often wants to intercept or modify, for water shaders, colored glass, and translucent-material lighting. OIT didn't just tweak a value: it replaced the shader files behind the old transparency pass with new ones (core/oit_composite.fsh and a set of oit_*.glsl includes) and removed the previous post-processing pipeline those files fed into. A pack that assumed the old pipeline's structure, rather than going through whatever abstraction Iris exposes, has a plausible reason to need updating for that alone.

That's reasoning about what's plausible, not a report of an actual broken pack, since none has been tested yet. If you use shaders, don't update your primary shader-enabled install on release day. Wait for Iris to publish a 26.3-tagged Modrinth release, then test your specific pack. The Iris shaders guide covers how Iris and Sodium fit together, and is the place to check once a 26.3 build lands.

If you use shaders, do not update to 26.3 the day it releases. Keep a 26.2 instance intact for your shader-enabled world, and only move it forward once Iris ships a 26.3 build and someone, ideally you, has actually confirmed your specific pack loads and renders correctly.

How this fits the rest of 26.3's rendering direction

OIT isn't an isolated change. It's the second piece of rendering-pipeline modernization in consecutive updates: 26.2 shipped an experimental Vulkan renderer as an opt-in alternative to OpenGL, and now 26.3 rewrites how translucency gets composited. Neither is a full rewrite on its own. Together they read as Mojang working through the renderer in stages, each piece getting a full snapshot cycle of testing before the next one starts.

For the rest of what 26.3 changes, including the mod and datapack compatibility picture, the what breaks in 26.3 guide covers that ground, and the full 26.3 feature rundown is the index for everything else in the update. If you're specifically chasing FPS or TPS problems that predate 26.3 entirely, the complete performance guide and the Sodium vs Embeddium comparison remain the baseline reference regardless of which Minecraft version you're on.

FAQ

What is OIT in Minecraft 26.3?

OIT stands for Order-Independent Transparency, a rendering technique that composites translucent surfaces (glass, water, ice, leaves) without needing to sort them by draw order first. It replaced the older Improved Transparency approach in Snapshot 2 (30 June 2026) and carried through unchanged to Pre-Release 1 (1 September 2026).

Does the Improved Transparency setting still exist in 26.3?

We could not find a definitive statement either way in the changelogs. What is confirmed is that the rendering approach behind Improved Transparency has been replaced by OIT, and the only documented toggle for OIT is a debug keybind (F3 + X), not a settings-menu option. That points toward OIT being unconditional now, but treat this specific point as our best reading of the evidence rather than a confirmed fact until checked against a Video Settings menu directly.

Does OIT lower FPS in Minecraft 26.3?

No public benchmark exists yet to answer this with a number, and this page will not invent one. Qualitatively, order-independent transparency techniques generally cost more per-pixel work than simple sorted blending, because they accumulate multiple translucent fragments instead of writing one. Any real-world cost would scale with how much translucent geometry is on screen: a little glass and water should barely register, a build made mostly of stained glass is where you'd expect to feel it most, if you feel it at all.

Will OIT break my shader pack?

Nobody can currently test this. Iris has no published 26.3 build as of 2 September 2026 (verified against Modrinth), so no shader pack has been run against 26.3's renderer yet. OIT did replace the shader files behind the old transparency pass, which gives shaders that hook into transparency a plausible reason to need updating, but that is reasoning about what's likely, not a confirmed break. Wait for an Iris 26.3 release before updating a shader-enabled install.

What does core/blit_depth.fsh in Pre-Release 1 actually do?

It's an alternative method for copying depth-texture data that OIT relies on, added because the original rendering-backend-based method "does not work on some devices," per the Pre-Release 1 changelog. Mojang didn't specify which devices. It's a compatibility fix for OIT itself, and a sign that the initial rollout had some hardware or driver combinations it didn't handle cleanly.

Should I update to 26.3 if I care about performance?

If you don't use shaders and your builds aren't unusually glass- or water-heavy, there is no verified reason to expect a meaningful FPS hit. If you use shaders, wait for an Iris 26.3 build before moving your shader-enabled install over. If you're on older or less-common GPU hardware, watch for transparency-related visual glitches after updating, since Pre-Release 1's blit_depth.fsh addition suggests OIT's original depth-copy method didn't work everywhere.

Sources & further reading: