C++29 mode is open, and the compilers already disagree

C++26 was finished in March. The next one is now something you can compile against.

LLVM 23.1.0 shipped on 25 August with -std=c++2d, and GCC’s development branch has the same mode. Neither current release will take the flag: clang 22.1 and GCC 16.2 both refuse it at the driver, before they ever look at your file.

Four lines

#include <cstdio>

int main() {
    std::printf("__cplusplus = %ld\n", (long)__cplusplus);
    return 0;
}

On GCC’s trunk in the new mode:

202700. Clang’s trunk in the same mode prints the same number, which is the useful part: two independent implementations picked the same placeholder for a standard that does not exist yet.

The convention is year and month of ratification, so 202700 is not a date anyone is committing to. It is a value chosen to sort after everything currently shipped, and it will be replaced by a real one when the standard is done.

The number they do not agree on

Ask the same two compilers about C++26 and the answer splits.

GCC’s trunk at -std=c++26 reports 202603. Clang’s trunk at -std=c++2c reports 202400:

202603 is the correct one. C++26 was completed in March 2026, so year-and-month gives 202603, and that is what GCC’s documentation of the predefined macros specifies. 202400 is the placeholder clang used while the standard was still in progress and has not yet updated.

This is a small thing that behaves like a large one. Feature-test macros are the recommended way to detect a feature, and for good reason, but __cplusplus is what a great deal of existing code actually tests, usually as __cplusplus >= 202302L or similar. A comparison written against the ratified value silently takes the wrong branch on clang, because 202400 < 202603 even though clang is in its C++26 mode. That will get fixed, and until it is, a version check spanning C++23 and C++26 is not portable between the two compilers.

Even the flag is not agreed

GCC accepts -std=c++29 and -std=c++2d, and gives 202700 for both. Clang accepts only -std=c++2d and rejects the spelled-out form:

error: invalid value 'c++29' in '-std=c++29'

The letter form is the older convention, 2a for what became C++20 and 2b for C++23, used while the year is unsettled. GCC adding the numeric alias early is a convenience rather than a claim about the schedule.

What you can actually do with it

Very little, deliberately. Turning the mode on does not hand you features. It opens a bucket for papers voted into the working draft after C++26, and the working draft has only just started collecting them. What you get today is a compiler that will not reject a paper’s syntax on the grounds that no mode enables it.

The reason to know the flag exists is narrower. If you follow a paper that has been voted in and want to try an implementation as it lands, this is where it will appear first. If you maintain a build matrix, this is the row that will start appearing in bug reports.

Sources: LLVM 23.1.0 release · Clang 23.1 release notes · GCC standard predefined macros · all four values measured on Compiler Explorer on 28 August 2026 · demo source in the examples repo.