Small Models that learn your business, runs it - and is yours.
Palo Alto
- Tweets 23
- Following 15
- Followers 395
- Likes 51
Introducing Drex - a lightning-fast decision model built to rival Jev.
#1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks
Architecture: Small Diffusion Model with RLAF
Price: $0.04 per 1M input tokens (cheaper than Jev)
Latency: less than a second.
Sign up now for 250M welcome credits.
nace.ai/drex
Open weights and the full tech report are coming very soon. 👀 🦖
#jev #nace #drex
23
34
281
449,568
Drex 1.1 is live - trained more with RLAF for in-code decision-making. More rivalry to Jev.
+3.1% on the Tools & Automation Index. Try both models depending on your use-case.
Claim 250M free welcome credits and start improving your agents: nace.ai/drex
Price: $0.04 per 1M tokens
Latency: less than a second
#drex #jev #nace
3
6
47
30,514
117-92. ggs
And there we go. I had a great time leading the training efforts for Drex, a diffusion based decision model. As a fun side benchmarking, We also evaluated the model on multiple board games from @kaggle Game Arena. It's scored 117-92 against Jev on the games played.
1
3
354
i replaced jev with a model x3 faster and totally FREE
250M free tokens to builders.
it's a decision model from Nace.AI — facts in, a probability for every option out. one pass, no long reasoning trace to parse.
full breakdown in this thread 👇
Introducing Drex - a lightning-fast decision model built to rival Jev.
#1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks
Architecture: Small Diffusion Model with RLAF
Price: $0.04 per 1M input tokens (cheaper than Jev)
Latency: less than a second.
Sign up now for 250M welcome credits.
nace.ai/drex
Open weights and the full tech report are coming very soon. 👀 🦖
#jev #nace #drex
19
11
91
11,830
Introducing Drex - a lightning-fast decision model built to rival Jev.
#1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks
Architecture: Small Diffusion Model with RLAF
Price: $0.04 per 1M input tokens (cheaper than Jev)
Latency: less than a second.
Sign up now for 250M welcome credits.
nace.ai/drex
Open weights and the full tech report are coming very soon. 👀 🦖
#jev #nace #drex
23
34
281
449,568
Nace AI retweeted
We will be dropping a decision model soon, surprisingly it's faster, more accurate than Jev on various benchmarks & yes, it beats Jev in 1v1 chess. ♟️
@kaggle @MeganRisdal Community would love to see Game Arena 2.0 w/ Decision Models. wdyt?
11
7
86
168,027
Releasing new research on HyperNetworks @ NaceAI Research Team. Introducing 📄Scaling Laws for Hypernetwork-Based Knowledge Injection in Large Language Models.
arXiv Preprint: arxiv.org/abs/2607.19604
More info: nace.ai/research/scaling-law…
3
8
33
17,182
🎉A warm welcome to the new Nacer 🚀
At NaceAI, our mission is build Long-Horizon AI that Runs Organizations.
Today, we take a massive step forward in that mission by welcoming @nischay_twt Nischay Dhankhar to the team as a Founding Research Engineer.
Nischay joins us with a pedigree that speaks for itself. A 3x Kaggle Grandmaster and the youngest person from India to achieve that status, Nischay currently sits in the Top 5 of the global rankings. His background in Computer Science and Machine Learning is complemented by a unique ability to translate complex deep learning theories into robust, production-ready systems.
As we move into our next growth chapter, Nischay will focus on:
Production post-training: Advancing the efficiency and reliability of our deployed models, and continuing our Test-Time Model Editing with Metamodels research directions to make AI more adaptable and precise.
Welcome to Nace AI, Nischay!
Join us in our mission! We’re hiring: nace.ai/careers 🎲
1
336
We're emerging from stealth with MetaModel 1, solving the critical challenge of real-time LLM customization. Finally, accurate, reliable, and policy-compliant Task Specific Language Models, tailored to your business
nace.ai/research/letter-from…
ALT Nace.AI
1
2
310