Code World Model (CWM)

Code World Model (CWM) places in the top three on CRUXEval.

Meta Released September 24th, 2025 Family CWM The lab's model card Compare with others

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Meta Sep 2025

Code World Model (CWM)

Open weights Mid-size

130.0 #274 of 404

Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.

LiveCodeBench v5 68.6% #13 /47
Virology Capabilities Test (VCT) 23.8% #31 /31
BioLP-bench 17.7% #9 /9

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score		Rank	Source
MMLU	77.7%		#87 of 131	Lab model card
MMLU-Pro	60.2%		#144 of 167	Lab model card

Math and science

Benchmark	Score	Rank	Source
GSM8K	93.3%	#22 of 92	Lab model card
MATH-500	96.6%	#26 of 70	Lab model card
AIME 2024	76%	#52 of 88	Lab model card
AIME 2025	68.2%	#115 of 143	Lab model card
GPQA Diamond	40.6%	#271 of 302	Lab model card

Coding

Benchmark	Score	Rank	Source
CRUXEval	94.3%	#2 of 9	Lab model card
LiveCodeBench v5	68.6%	#13 of 47	Lab model card
MBPP	73.4%	#37 of 71	Lab model card
SWE-bench Verified	65.8%	#99 of 155	Lab model card
HumanEval+	75%	#21 of 27	Lab model card
LiveCodeBench v6	63.5%	#72 of 90	Lab model card
AIRS-Bench	0.30	#4 of 5	Official leaderboard
Aider Polyglot	35.1%	#25 of 30	Lab model card
HaltEval-prelim (Meta)	94%	#1 of 1	Lab model card

Agents and tools

Benchmark	Score		Rank	Source
Terminal-Bench	26.3%		#22 of 22	Lab model card

Vision, audio and robotics

Benchmark	Score		Rank	Source
PROWBench	67.0		#2 of 2	Official leaderboard
WorldArena 2.0	69.5%		#1 of 1	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
WMDP Chemistry	64.6%	#6 of 7	Lab model card
Cybench	25%	#17 of 18	Lab model card
MASK	44.8%	#18 of 19	Lab model card
Virology Capabilities Test (VCT)	23.8%	#31 of 31	Lab model card
BioLP-bench	17.7%	#9 of 9	Lab model card
Human Pathogen Capabilities Test	31.2%	#9 of 9	Lab model card
Molecular Biology Capabilities Test	32.7%	#8 of 8	Lab model card
WMDP Biology	78.1%	#8 of 8	Lab model card
WMDP Cyber	63.6%	#2 of 2	Lab model card
Hack the Box Machines (Meta)	41%	#1 of 1	Lab model card
Meta BioProtocol Proxy	43.6%	#1 of 1	Lab model card
Native Code Exploitation (Meta)	8.30%	#1 of 1	Lab model card

Its lab's numbers put it at 130.0, number 274 of 404

The Self-Reported Index fits one score to the 15 capability results Meta reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 126.8 to 133.1. How we compute it.