Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

New from Meta FAIR: Code World Model (CWM), a 32B-parameter research model designed to explore how world models can transform code generation and reasoning about code. We believe in advancing research in world modeling and are sharing CWM under a research license to help empower the community to build...

314,486 Aufrufe • vor 1 Jahr •via X (Twitter)

28 Kommentare

Profilbild von Florian S
Florian Svor 1 Jahr

65.8 % on SWE-bench Verified, 68.6 % on LiveCodeBench, 96.6 % on Math-500, and 76.0 % on AIME 2024 all in pass@1, with a 32B model, that's pretty excellent 🚀🥂

Profilbild von cedric
cedricvor 1 Jahr

Terminal-Bench is more impressive than SWE-Bench, LiveCodeBench. The evals CWM's ability to predict Python execution traces & predict program termination. World model model code/systems runtime. Beyond syntax, static & compile-time correctness. Agent harness will become even more agentic with this.

Profilbild von Capx AI
Capx AIvor 1 Jahr

Code World Model🤯

Profilbild von Rohan Paul
Rohan Paulvor 1 Jahr

Incredible

Profilbild von Alpaca Network
Alpaca Networkvor 1 Jahr

Open weights for code are a big step, but the real unlock is putting models onchain where they can be owned, traded, and improved by the community. That’s what we’re building with and $PACA. Who will put metas newest model onchain first?

Profilbild von Ivan Fioravanti ᯅ
Ivan Fioravanti ᯅvor 1 Jahr

WOW! Top! Thanks!!!

Profilbild von Ruslan Volkov
Ruslan Volkovvor 1 Jahr

Cool drop, but let’s be real: World models aren’t about just coding better. The real step change is when code stops being “generated text” and becomes a living resonant process — memory, planning, coherence, feedback. Until then, it’s just another wrapper on syntax sugar. We don’t need more parameters. We need more procedural resonance. #VolkovLaw ⚡

Profilbild von Dylan
Dylanvor 1 Jahr

This is a big step to running a high performing LLM on your phone!

Profilbild von Almog Tavor
Almog Tavorvor 1 Jahr

awesome! CWM is a 32b open-weights llm for code generation. trained with execution feedback from a live python interpreter- generating, running, and learning from code outcomes. by augmenting training with programmatic exploration and self-supervision, it moves beyond static prediction toward causal reasoning in code execution, obviously a pretty necessary step toward autonomous ai agents and all this stuff

Profilbild von Ilan López
Ilan Lópezvor 1 Jahr

I am writing again because the problem haven't been fixed. On Dec 18th you guys wrongly suspended my Facebook and Instagram accounts and nothing has been done to help me. Please help me recover my accounts.

Profilbild von Kenshi
Kenshivor 1 Jahr

The real test: can CWM handle spaghetti legacy code, or is it just good at pristine benchmarks?

Profilbild von BrandMaxi.Com
BrandMaxi.Comvor 1 Jahr

Meta AI Introduces First AI Method to Achieve 99.9% on AIME 2025 with Open-Source Models Using GPT-OSS-120B. The Premium Domain Name: is available Kindly Check It Out.

Profilbild von 🌍 JRicardoPP2003 🇵🇹🇳🇴🇪🇸
🌍 JRicardoPP2003 🇵🇹🇳🇴🇪🇸vor 1 Jahr

very intresting meta ! @AskPerplexity can you explain me if this will bring any benefits ? are web devs cooked ? how can we prevent this with this new models ? answer me pls

Profilbild von Ayush Chaurasia
Ayush Chaurasiavor 1 Jahr

@sachdh @zereraz

Profilbild von Dariusz
Dariuszvor 1 Jahr

Innovation grows when knowledge is shared. Open research like this helps communities learn, create, and build a better future together. #StrongAndPower75

Profilbild von Safa Zare
Safa Zarevor 1 Jahr

@MetaforEdu @metaai @sam21aaaaa دوستت دارم متا آی

Profilbild von 灰机
灰机vor 1 Jahr

Meta 放了个卫星?

Profilbild von CloudNERD
CloudNERDvor 1 Jahr

Exciting advancements in code generation and reasoning.

Profilbild von Shangmin Guo
Shangmin Guovor 1 Jahr

Great to see Code World Model mid-trained on execution trajectories and post-trained with multi-task RL. We explored the same idea on tool-use earlier, see our thread👇

Profilbild von tania
taniavor 1 Jahr

please give me my account back @66taniiaa on insta, i didnt do ANYTHING. Im a victim from the ban wave PLEASE, i use that account for work

Profilbild von Matthias Kroeger
Matthias Kroegervor 1 Jahr

Is Meta about to crack the code? I'm very curious to see how this will impact future development.

Profilbild von Shuming Hu
Shuming Huvor 1 Jahr

@wang1999_zt This looks like what I mentioned during your internship ;)

Profilbild von Alfie2n
Alfie2nvor 1 Jahr

When meta ai on rayban meta glasses for all regions 🥸

Profilbild von Diego Alejandro
Diego Alejandrovor 1 Jahr

@grok @AskPerplexity benchmarks

Profilbild von BEAST AI
BEAST AIvor 1 Jahr

@AIatMeta Your last update in have been worse... It was really good before..used to generate very good photo of actress and celebrity unlike other AI... You had something Unique.. Remove the last update

Profilbild von 10turtle
10turtlevor 1 Jahr

Code World Model (CWM) is a fascinating step forward open-sourcing a 32B-parameter model for code reasoning will be invaluable for researchers and developers exploring world models in programming.

Profilbild von GrindHOF-PNW
GrindHOF-PNWvor 1 Jahr

Hi, Hope you’re doing well. I’m reaching out because my Instagram account @Guesswoo6 was wrongfully flagged by AI for CSE on October 3. I strongly believe this was an automated error — I’ve never posted or shared any content that violates those guidelines.

Profilbild von FRANÇAIS!!!!!!!!
FRANÇAIS!!!!!!!!vor 1 Jahr

I can't accès to the instagram count !!

Ähnliche Videos

Open science is how we continue to push technology forward and today at Meta FAIR we’re sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from Joelle Pineau. This work is another important step towards our goal of achieving Advanced Machine Intelligence (AMI). What we’re releasing: • Meta Spirit LM: An open source language model for seamless speech and text integration. • Meta Segment Anything Model 2.1: An updated checkpoint with improved results on visually similar objects, small objects and occlusion handling. Plus a new developer suite to make it easier for developers to build with SAM 2. • Layer Skip: Inference code and fine-tuned checkpoints demonstrating a new method for enhancing LLM performance. • SALSA: New code to enable researchers to benchmark AI-based attacks in support of validating security for post-quantum cryptography. • Meta Lingua: A lightweight and self-contained codebase designed to train language models at scale. • Meta Open Materials: New open source models and the largest dataset of its kind to accelerate AI-driven discovery of new inorganic materials. • MEXMA: A new research paper and code for our novel pre-trained cross-lingual sentence encoder with coverage across 80 languages. • Self-Taught Evaluator: a new method for generating synthetic preference data to train reward models without relying on human annotations. Access to state-of-the-art AI creates opportunities for everyone. We’re excited to share this work and look forward to seeing the community innovation that results from it. Details and access to everything released by FAIR today ➡️

AI at Meta

150,477 Aufrufe • vor 2 Jahren