The Short Answer
Quick answer: Stockfish is rated about 3650 on CCRL's 40/15 list, the main rating list for chess engines. On CCRL's blitz list, Stockfish 19 ranks first at 3804. That's far above Magnus Carlsen's FIDE rating of 2823, but the two scales never meet, so the true gap isn't known.
I pulled both CCRL lists on 31 October 2026. I also read the part of Stockfish's code that sets its weaker playing levels.
Below is what each number means, how much each version added, and what the levels are worth.
What I Checked
Engine ratings often get quoted with no date and no source. So here's where every number in this guide comes from.
- CCRL's 40/15 list, computed on 30 October 2026 from 2,477,101 games.
- Its blitz list, computed on 27 October 2026 from 2,164,740 games.
- Release notes for Stockfish 18 and 19, on GitHub.
- The source code that turns a chosen Elo into a skill level, plus the 2023 code change behind it.
- Wikipedia's page on Stockfish, for its history and tournament record.
CCRL stands for Computer Chess Rating Lists. Its volunteers run engine against engine games at fixed time controls and rate the results. I compared each Stockfish version across both lists.
Stockfish's Rating on Each List
On the CCRL 40/15 list, Stockfish 18 on four processor cores rates 3649, give or take 12 points, over 1,466 games. Stockfish 19 on four cores also rates 3649, over 738 games, with a margin of 16.
First place belongs to Triumviratus 8.0, at 3650. It has played only 214 games, though, so its margin is 28 points either way. With that margin, you should read a one point lead as a tie.
The time control equals 40 moves in 15 minutes on an Intel i7-4770k desktop processor. Engines play with pondering off, from a shared opening book of up to 12 moves. Wikipedia rounds Stockfish's figure to 3650.
On the CCRL blitz list, Stockfish 19 on eight cores ranks first at 3804. Triumviratus 7.0 follows at 3780, and Torch v4 at 3778. Blitz there means two minutes plus a one second increment, on the same processor.
Why the Blitz Number Is Higher
The 3804 doesn't mean Stockfish plays better at blitz than at slower chess. Each list rates its own pool of engines, on its own scale.
A rating only measures results against the players in the same pool. The blitz list has different engines, a different number of games and a different time control. So the gap between the two lists isn't a strength gap.
If you compare engines, compare numbers inside one list only. On the 40/15 list, Stockfish sits one point behind first place. On the blitz list, it leads by 24 points.
Can You Compare It With Carlsen?
Not directly. Magnus Carlsen is rated 2823 on FIDE's October 2026 list. Stockfish's 3650 comes from games against other engines only.
Elo measures results inside one pool, as our Elo system guide explains. Humans and engines don't play rated games against each other, so no list joins the two pools.
That's why "Stockfish is 800 points stronger than Carlsen" isn't a measured figure. The real gap could be larger or smaller.
What's clear is the result. The Stockfish team's notes for version 18 say it "is stronger than any human, even when running on older or low-end hardware."
The 2014 Handicap Match With Nakamura
The closest thing to a human test came in August 2014. Hikaru Nakamura, then rated 2798, played Stockfish 5 in a four game handicap match, according to Wikipedia.
- In two games, Nakamura had help from an older version of the Rybka engine.
- For the other two, he got pawn odds and no help.
- Neither half let Stockfish use its opening book or endgame tablebases.
Stockfish won each half 1.5 to 0.5. Today, Stockfish 5 rates 3238 on the 40/15 list. Version 19 on one core rates 3629, almost 400 points higher.
Stockfish's Elo Over Time
CCRL keeps old versions on its list, so you can trace the climb. These are single core ratings from the 40/15 list.
- Release 1.01, one of the earliest, rates 2740. Stockfish 1.0 came out in November 2008.
- By version 3, the rating reached 3129.
- Version 5, the one that played Nakamura, rates 3238.
- Then came version 8, which AlphaZero beat in December 2017, at 3358.
- NNUE, a fast neural network for judging positions, arrived with version 12. That version rates 3543.
- Versions 16 and 17 both rate 3618.
- Today's version 19 rates 3629.
From version 8 to version 12, the rating rose 185 points. From version 12 to version 19, it rose 86. Our AlphaZero match story covers what happened in 2017.
What Each New Version Claims
The release notes give bigger numbers. Version 18 brought "an Elo gain of up to 46 points" over version 17, the team wrote. Version 19 claims "up to 44 points" over version 18.
The Stockfish 19 notes add that it "wins more than three times as many game pairs as it loses" against version 18.
Yet the CCRL lists show smaller steps. On the 40/15 list, version 19 on one core rates 3629, one point above version 18. On the blitz list, the gap is 21 points.
So which number should you trust? Both, for different questions. The team's tests pit the new version against the old one directly, in paired games. CCRL rates each version against many other engines, and most of those games end in a draw.
Why Most Top Games End in a Draw
Draws help explain why the top engine ratings bunch together. Across the whole 40/15 list, 51.1% of games end drawn. For Stockfish 19 on four cores, it's 92.8%.
A draw against a lower rated engine costs rating points. So a strong engine needs wins to pull ahead, and wins are rare at the top. Stockfish 19 scored 53.5% on that list, against opponents rated about 21 points lower on average.
Blitz games end drawn less often. On the blitz list, 46.6% of all games are draws. Stockfish 19 on eight cores drew 89.5% of its blitz games.
Hardware Changes the Number
More processor cores make Stockfish stronger, and CCRL rates each setup on its own line. On the 40/15 list, Stockfish 19 rates 3629 on one core and 3649 on four. On the blitz list, it rates 3792 on one core and 3804 on eight.
So the number you quote depends on the machine. A phone, a laptop and a 64 core server each run a different strength of the same program.
Version 18's notes put a figure on the top end. On the best hardware, it searches "over 500 million positions per second."
Stockfish's Skill Levels and Elo Setting
Stockfish can play weaker on purpose. Its settings take a skill level from 0 to 20, or a target Elo that the code turns into a skill level.
Its source code says skill levels 0 to 19 cover "CCRL Blitz Elo from 1320 to 3190, approximately." Level 20 is full strength. A January 2023 code change set those numbers, from games between Stockfish at each level and versions of a weaker engine called Stash, which CCRL rates.
In that test, level 0 came out at 1320 and level 5 at 2204. Level 10 reached 2788, and level 19 reached 3191. The same note warns that the scale is only accurate to within about 100 points, because results depend on the opponents.
So if you set Stockfish to 1500, it plays at 1500 on an engine scale. It doesn't have to match a 1500 player on Chess.com or in FIDE events.
Stockfish Levels on Lichess
Lichess's computer opponent runs Stockfish at eight levels. Each level sets a skill level, a think time and a search depth, not a rating.
Level 1 thinks for about 50 milliseconds at the lowest skill. Level 8 plays at full skill, with about one second per move and a search depth of 22.
That's why a search for "Stockfish level 8 Elo" finds only guesses. Lichess sets each level by skill and time, so a level has no fixed Elo.
Tournaments and the Testing Behind It
Stockfish's record in engine events matches its ratings. Wikipedia says it has won the Top Chess Engine Championship, or TCEC, 19 times. It has also won every main TCEC and Chess.com Computer Chess Championship event since 2020.
Fishtest drives its progress. Volunteers donate computer time, and a code change stays only if it beats an older reference version over many games. After Fishtest began in 2013, Stockfish gained 120 Elo in 12 months, Wikipedia says. By October 2026, the system had used more than 20,700 years of processor time on over 10.3 billion games.
Its rivals take other routes. Our Stockfish vs Leela vs Komodo comparison covers the main ones. Newer engines such as Torch and Triumviratus now sit within a few points of it on CCRL.
What the Number Means for Your Games
For your own games, the exact Elo doesn't matter much. Even on modest hardware, any recent Stockfish outplays every human, by its team's own account.
Use it as an analysis tool instead. Its evaluation score shows who stands better, and our evaluation score guide explains how to read it.
As an opponent, full strength Stockfish teaches little, because you'll almost surely lose every game. A lower skill level helps more. Practice against our chess bot, then check your mistakes with an engine afterward.