Our Models

Explore our family of multilingual language models designed for enhanced reasoning, reflection, and cross-cultural communication.

A model family based on Qwen3 trained for 100+ languages, finetuned for logic, truthfulness, and reflection.

GPRO-aligned

Unsloth + QLoRA

T4/4090-friendly

Stepwise Reasoning

Multilingual

Open-Source

Variant	Parameters	Languages	Specialization	Links
Noema-T1-4B	4B	100+	Local Deployment	HF GitHub
Noema-T1-8B	8B	100+	Enhanced reasoning	HF GitHub
Noema-T1-14B	14B	100+	Scalable AI model	HF GitHub
Noema-T1-32B	32B	100+	Perfect for research and power	HF GitHub

Model Roadmap

Initial release of our multilingual model family with reasoning capabilities.

Expanded language coverage and enhanced reasoning capabilities.

Development of domain-specific models for scientific reasoning and cross-cultural communication.

Exploration of multimodal reasoning across languages and cultures.