CFOP Tutorial — Speedsolve the Rubik's Cube
What is CFOP
Before you start: this page assumes you can already solve the cube with Layer-by-Layer (LBL) — you know faces, edges, corners and layers, you understand "build up one layer at a time without breaking what's done," and you've been through the minimal notation in Fundamentals. CFOP builds on the LBL you already know; it doesn't repeat those basics.
CFOP reorganizes LBL's steps into four: Cross, F2L (first two layers), OLL (orient the last layer), PLL (permute the last layer). Each step maps to an LBL move you already know — it just bundles several of them into fewer, bigger steps. What you get out is the same solved cube.
If LBL is "do one small piece per step," CFOP is "do one big piece per step." It's for someone who can already solve and wants to be faster: same destination, fewer moves, smoother flow. The cost is more to memorize — OLL and PLL each have an algorithm set to learn. The trade-off in one line: you trade memory for speed.
How do CFOP's four steps come out of LBL? Why does merging them save moves? The next section covers the core idea.
Core idea: merge LBL's steps
You already solve with LBL: build the bottom cross, fill the bottom corners, do the middle-layer edges, then finish the top in a few sub-steps. CFOP does exactly the same thing — it just merges some of those steps. F2L fuses "bottom corners" and "middle edges" into one step, inserting a corner and its paired edge into a bottom slot at the same time; OLL merges the top "orientation" work into a single pass; PLL merges the top "positioning" work into a single pass. Fewer steps, fewer turns, smoother flow. That's the whole of CFOP relative to LBL: not a new method, but your existing steps repackaged.
CFOP's cross goes on the bottom (same place as your LBL cross) — and that's not just a position choice, it's staging the real source of CFOP's speed: lookahead. Lookahead means while you execute the current pair, your eyes are already finding the next, so pair follows pair without a pause. Most solve time is actually lost to stopping to find the next piece, not to the turns themselves. CFOP's structure is built for this — the cross pressed underneath keeps the finished parts hidden and untouched, while the unsolved top stays facing up and in view, so the next pair is always in sight. But remember: fewer moves alone don't make you faster; the moves you save only become real time saved once lookahead catches them and turns them into unbroken flow.
CFOP vs LBL — what actually changes, and is it worth it
The difference fits one mapping: CFOP's Cross is exactly LBL's cross (identical); F2L merges LBL's "bottom corners + middle edges" into one step; OLL merges "top cross + top corner orientation" into one pass; PLL merges "top corner placement + top edge placement" into one pass. Four steps absorb three of LBL's splits, so there are fewer turns and shorter transitions — that's the real source of its speed, independent of how fast your hands move.
But speed has a price. LBL finishes the top with a few reusable small moves — repeated, never memorized. To do it "in one pass," CFOP's OLL and PLL each keep a full algorithm set — roughly 57 for OLL, 21 for PLL. You trade "more to memorize" for "done in one look": every case has a ready-made shortcut that never re-does work. So CFOP isn't free speed — it's a deliberate trade.
Two realities to know: first, F2L is the hardest part of CFOP — "watch a pair, insert together" takes more thinking than LBL's two separate steps, so at first you'll be slower than with LBL; second, the algorithms aren't learned overnight — most people start with "2-look" OLL/PLL (a couple of algorithms each) to get the whole solve working, then gradually expand to the full sets. Learning CFOP is a "slower before faster" road.
Roadmap
Below are CFOP's four steps, in solve order, top to bottom. Click any step to jump to its concept intro further down.
Phase tour: four steps that chain into one story
The four steps chain in order: each follows the last and points to the next. Read this section and you'll understand CFOP end to end — and how each step maps back to LBL.
Cross: the same thing as LBL's cross — send the four white edges to the bottom, white facing down, side colors matched to the adjacent centers, making the white cross on the bottom face. Where CFOP differs is how: rather than "place then adjust," you aim to place all four at once, in as few turns as possible (a good cross takes about seven or eight moves). With the cross on the bottom, it sits underneath and stays out of the way while you do F2L next.
F2L (first two layers): this is where CFOP and LBL part ways. LBL does the four bottom corners and the four middle edges as two separate steps; F2L merges them — for each of the four bottom slots (front-left, front-right, back-left, back-right), you take a corner and its paired edge and insert them as one unit into the slot. Once all four pairs are in, the bottom two layers are done together. This step is the core of CFOP's move savings — and the hardest to pick up. With two layers complete, only the top remains — orient the whole top face in one pass next (OLL).
Why insert the corner and edge together
In LBL, a bottom corner and a middle edge are inserted separately — corner first, then edge, sometimes getting in each other's way. F2L's insight: a bottom corner and the middle edge beside it belong in the same slot anyway, so they can be handled as a pair — pair them up in the top layer first, then slide both into the slot together. One move resolves a corner and an edge at once, skipping the back-and-forth of "do the corner, then come back for the edge." That's the essence of merging to save moves.
OLL (orient the last layer): flip the orientation of every sticker on the top layer in one pass, turning the top face a uniform yellow (LBL splits this into "top cross" and "top corner orientation," two sub-steps; CFOP merges it into one). CFOP pulls this off with a full set of OLL algorithms — this is where "you trade memory for speed" pays off. This step only cares which way colors face, not where pieces sit; the bottom two layers stay untouched. Once the whole top is yellow, orientation is done — what's left is moving those top pieces into place (PLL).
PLL (permute the last layer): the final step — move every top piece, already oriented, into its correct position (LBL splits this into "top corner placement" and "top edge placement," two sub-steps; CFOP merges it into one). It too relies on a full set of PLL algorithms, done in one look. Once everything is placed, all six faces are complete and the cube is solved — the finish line of CFOP.
How to learn it
Every phase follows the same loop: read its concept on this page first, then open the phase's standalone tutorial for its structure (how it breaks into cases, how to recognize them), then head to the practice area to drill that phase's cases by hand — and only move on once it's solid.
Go in order: Cross → F2L → OLL → PLL — all four depend on each other layer by layer; skip one and there's no "finished part" for the next to lean on. But the two top steps (OLL, PLL) don't need the full sets up front: start with "2-look" OLL/PLL (a couple of algorithms each) to get the whole solve working end to end, then gradually expand to the full sets.
Be prepared: when you start F2L you'll be slower than with LBL — "handle a pair at once" takes more thinking than "two separate steps," and only flips into a net win once it's fluent. A soft bar for "got it": without the tutorial, you can work most cases and do them steadily — then move on. F2L is the watershed of the whole route; it deserves the most time.
Finally, remember the split across three layers: the method page (this one) is a map you read once, to understand CFOP end to end and how it maps to LBL; the phase tutorial goes deep on one step's structure; the practice area is where you drill cases repeatedly. Use all three together. Curious about the theoretical ceiling — how a computer solves any cube in ~20 moves? See Kociemba (near-optimal solving).