That's why you need DatologyAI

Agree. Overflowing with integrity

The team absolutely crushed it here. They blew away nearly every CLIP baseline, and matched or exceeded SigLIP2—which uses a slew of training algorithm improvements—on a number of benchmarks. USING. DATA. CURATION. ONLY. I’m so proud of Ricardo Monti , Haoli Yin ,

Datology CLIP Models DatologyAI releases two SOTA CLIP ViT-B/32 variants: classification-optimized and retrieval-optimized, achieving top results through task-specific data curation alone. Model - ViT-B/32 (86M params), OpenCLIP 2.24.0 - No architecture or training changes -

This is actually one of the less insane things we've done

👀 Thrilled to be part of the DatologyAI team! Can't wait to start contributing to our mission of high quality data curation!

So excited to have you!!

Congratulations to our friends and partners Arcee.ai on the release of AFM-4.5B! With data powered by DatologyAI, this model outperforms Gemma3-4B and is competitive with Qwen3-4B despite being trained on a fraction of the data.

AFM - Arcee Foundation Models. Built from scratch for enterprise. The first release, AFM-4.5B, is a 4.5B open-weight model that runs anywhere: cloud, edge, or CPU. Trained on rigorously filtered data with full deployment flexibility. I don’t say this lightly ⮕ 𝑫𝑶𝑵’𝑻

We've definitely seen signs of this already — perhaps not surprisingly, post-training people tend to care more about the value of data. We see a number of companies turning to DatologyAI for getting the most out of their existing datasets!

One of the dreams when joining DatologyAI was to bring the fruits of data research from labs🔬 to the real world 🌎 Soo gratifying to see that our algorithms are out in the open, enabling companies to rival the sophisticated Qwen & Gemma families at a fraction of the cost!

Partnering w/ Arcee.ai was a blast. This is the first public language model pretrained on DatologyAI-curated data, and we're pleased (though not surprised) that it goes toe-to-toe w/ the best small models. The base model is 🔥 and Arcee's post-training expertise (S-tier) really

It's crazy what you can do with good data It's crazy what you can do with DatologyAI

kalomaze Aside from the "big guys" (Gemma3, Llama 3.3, Qwen-MAX), Arcee models have ALWAYS, ALWAYS been the best in terms of not being lobotomites. I know kalomaze works with Prime Intellect (who has used Arcee)…but people need to understand, Arcee is the GOAT of posttraining. And their

"Pretraining is dead" is dead

At DatologyAI, every employee is employee of the month! Every month!

So excited to have you, Sid!

You think you're a data researcher? Name C4's 10 worst samples.