Riemann Computing Inc. Creates and Demos Out a 13B Parameter AI Model, Aims at Massively Reducing Training Costs

Riemann Computing Banner Image

Riemann Computing Founder on GoFundYourself!

Riemann Computing Utilizes the NTK Training Technique Over LoRa for Cost-Effective AI Model Training

SAN FRANCISCO, CA, UNITED STATES, October 21, 2026 /EINPresswire.com/ -- Riemann Computing, the company behind the OpenPeer AI family of models, announced the prerelease of My Name Is Riemann, a 13.1-billion-parameter decoder-only language model. The release marks the first public build in the My Name Is Riemann line and represents a major step in the company’s effort to develop scalable, openly accessible AI systems.

My Name Is Riemann was grown “rung by rung” from the published weights of OpenPeerLLM, Riemann Computing’s foundational OpenPeer AI model. Rather than using conventional low-rank adaptation methods such as LoRA, the model was fine-tuned with the company’s OpenPeer-AI-NTK-Trainer gate controller. Riemann Computing says this approach reflects its broader strategy of building AI infrastructure around modular, distributed, and community-accessible machine learning.

The newly released model contains 13,110,702,080 parameters and is distributed in FP32 precision across eleven safetensors shards, totaling 52.4 GB. The model features a hidden size of 5,120, 40 transformer layers, 40 attention heads, and a feed-forward network size of 20,480. It supports a 2,048-token context window.

The final 13.1B stage was trained on 2.16 billion tokens drawn from a 2.30-billion-token corpus. Across all six stages of its development, the model has received 6.5 billion cumulative training tokens. Riemann Computing characterizes the token budget as deliberately modest relative to the model’s size. The prerelease was trained between October 19 and October 20, 2026, and is now available as a complete, loadable model on the Hugging Face Model Hub under the OpenPeerAI organization.

By publishing the model publicly, the company invites researchers, developers, and independent builders to inspect, test, and experiment with the system before its next stages of training.

The OpenPeer AI initiative, which began with OpenPeerLLM, is intended to provide a scalable contextual base for language modeling, grid-based learning, and contextual text generation. The initiative has described its mission as bringing distributed, safe, and innovative machine learning to a wider audience. “My Name Is Riemann” extends that foundation into a substantially larger parameter class. Riemann Computing says the prerelease gives the public an early look at how its staged growth process performs as models scale beyond the original OpenPeerLLM base.

Riemann Computing emphasizes that this is a prerelease rather than a finished production model. Future versions are expected to receive additional training and refinement as the company continues its staged development process. The release also arrives as Riemann Computing maintains an active public fundraising presence on Wefunder, where it has shared updates on its broader technology roadmap.

The company says the prerelease will help it gather real-world feedback while continuing work toward more capable future versions of the My Name Is Riemann model family.

As always, none of this is investment or financial advice. Please do your own due diligence and research.

Andrew Kamal
Riemann Computing Inc.
+1 248-238-8245
Visit us on social media:
LinkedIn
Instagram
YouTube
X