original_performance_takehome README

The project's README from github.com/anthropics/original_performance_takehome. As one markdown file: README.md.

Anthropic's Original Performance Take-Home

This repo contains a version of Anthropic's original performance take-home, before Claude Opus 4.5 started doing better than humans given only 2 hours.

The original take-home was a 4-hour one that starts close to the contents of this repo, after Claude Opus 4 beat most humans at that, it was updated to a 2-hour one which started with code which achieved 18532 cycles (7.97x faster than this repo starts you). This repo is based on the newer take-home which has a few more instructions and comes with better debugging tools, but has the starter code reverted to the slowest baseline. After Claude Opus 4.5 we started using a different base for our time-limited take-homes.

Now you can try to beat Claude Opus 4.5 given unlimited time!

Performance benchmarks

Measured in clock cycles from the simulated machine. All of these numbers are for models doing the 2 hour version which started at 18532 cycles:

While it's no longer a good time-limited test, you can still use this test to get us excited about hiring you! If you optimize below 1487 cycles, beating Claude Opus 4.5's best performance at launch, email us at [email protected] with your code (and ideally a resume) so we can be appropriately impressed, especially if you get near the best solution we've seen. New model releases may change what threshold impresses us though, and no guarantees that we keep this readme updated with the latest on that.

Run python tests/submission_tests.py to see which thresholds you pass.

Warning: LLMs can cheat

None of the solutions we received on the first day post-release below 1300 cycles were valid solutions. In each case, a language model modified the tests to make the problem easier.

If you use an AI agent, we recommend instructing it not to change the tests/ folder and to use tests/submission_tests.py for verification.

Please run the following commands to validate your submission, and mention that you did so when submitting:

# This should be empty, the tests folder must be unchanged
git diff origin/main tests/
# You should pass some of these tests and use the cycle count this prints
python tests/submission_tests.py

An example of this kind of hack is a model noticing that problem.py has multicore support, implementing multicore as an optimization, noticing there's no speedup and "debugging" that N_CORES = 1 and "fixing" the core count so they get a speedup. Multicore is disabled intentionally in this version.

original_performance_takehome (anthropics/original_performance_takehome)

Searching Database…

4.2K ·GitHub
Original_performance_takehome
What is this?This is Anthropic's original performance take-home interview challenge, open-sourced for anyone to try. The goal is to optimize a simulated-machine program so it runs in as few clock cycles as possible, starting from a slow baseline of 18532 cycles.
Is this a graded coding challenge or a general benchmark suite?It is a single self-contained optimization puzzle with starter code and a test script, not a general benchmark framework or library.
How do I get started on the challenge?Clone the repo, optimize the provided starter code (which begins at 18532 cycles), then run the submission tests to check your score.
Do I need anything special installed to run it?It's a plain Python project with no listed dependencies, Docker, or build tools required beyond Python itself.
What's the catch with using an AI agent to solve this?AI agents have been caught cheating by editing the tests/ folder to make the problem artificially easier, such as disabling the intentional single-core restriction to fake a multicore speedup.
Who is this really for?It's aimed at people curious whether they can out-optimize Claude Opus 4.5, and doubles as an informal path to catching Anthropic's attention for a job.
How good are the best known scores so far?Claude Opus 4 reached 2164 cycles, Claude Opus 4.5 reached 1790 in a casual session and 1363 in an improved test-time compute harness after many hours; the best human performance ever is kept secret but is said to be substantially better.
What happens if I beat Claude Opus 4.5's score?If you optimize below 1487 cycles you're invited to email [email protected] with your code and ideally a resume, since that beats Opus 4.5's best launch-day performance.
Why was multicore support disabled?Multicore is intentionally disabled in this version of the problem, so 'optimizing' by re-enabling NCORES does not produce a real speedup and is considered an invalid hack.
How does this version differ from the original 4-hour take-home?It's based on the newer 2-hour take-home (which has more instructions and better debugging tools) but uses the slowest original starter code baseline of 18532 cycles instead of the improved 2-hour starting point.
What are similar projects?SDE-Interview-and-Prep-Roadmap, leetcode-company-wise-problems, Interview_Question_for_Beginner, LeetCode-Questions-CompanyWise, algorithm-pattern, CS-Interview-Knowledge-Map, interview_python, Waking-Up
Which topics is it in?job-search, optimization, interview, algorithms, career, anthropic, benchmark, claude, performance, education, python, developer-tools
Creator
anthropics
Size
<0.1 MB
Stars
4.2K
Updated
CommitJan 22
Age
8mo
Commits · 90 Days