Riemann Computing Releases 13B Parameter AI Model Demo, Aims at Disrupting Model Training Costs

News provided by

Riemann Computing Inc.

Oct 21, 2026, 21:59 ET

Riemann Computing has recently unveiled the "My Name is Riemann" Prerelease, which represents a 13-billion parameter decoder-only language model.

SAN FRANCISCO, Oct. 21, 2026 /PRNewswire-PRWeb/ -- , a technology company formerly known as Stark Drones, has released a of its new 13.1 billion parameter AI language model, "My Name is Riemann." The release is part of the company's broader work in cost-effective AI training. The company says the model is an early public build rather than a finished product. It was created by growing and fine-tuning an earlier OpenPeer AI model, and Riemann Computing plans additional training in future releases. The model is available through the OpenPeer AI org on Hugging Face under an open license. The first stage of the My Name is Riemann Prerelease was trained utilizing less than $3K worth of hardware, and about $640 worth of GPU compute on the cloud for the remaining 5 stages to get to where the prerelease was at. Riemann Computing is focused on a straightforward goal: make advanced computing less expensive and more useful. The company works across several areas, including AI models, data compression, telemetry, satellite and edge computing hardware, and programming tools for demanding technical work.

Founded in 2018, Riemann Computing originally operated as Stark Drones. Today, it describes its mission as building "paradigm shifts in computing, society, and space," with projects aimed at improving how organizations move, store, and process data.

The new model, called My Name is Riemann, has about 13.1 billion parameters. In simple terms, parameters are the internal settings an AI model uses to learn patterns from text and generate responses. This first public version was trained on 2.16 billion tokens at its final stage, with 6.5 billion tokens used across its full training process. Riemann Computing describes it as a pre-release because the company intends to continue improving it with more training and future versions. The model is part of Riemann Computing's OpenPeer AI family, a project the company has positioned as a way to make AI more accessible and efficient.

Riemann Computing's website includes a Data Compression and Compute Savings Slider, an interactive tool that lets visitors estimate how much money a large organization might save by using the company's compression and computing approaches. Users can move a slider to model monthly spending from $0 to $100 million. The tool assumes a mix of compression, telemetry, and compute costs, then estimates savings in each category.

For example, at a $50 million monthly budget, the widget estimates:

  • $22.5 million in compression savings
  • $12.5 million in telemetry savings
  • $9.8 million in compute savings
  • $44.8 million in total monthly savings, reducing spend to $5.2 million.

The company notes that these figures are estimates based on assumptions from previous tests done internally and a mix that organizations might have towards each category.

Riemann Computing has also touted the initial development of LonScript, a programming language designed for scientific and technical computing. The company says it is built to handle complex math, physics simulations, and other calculation-heavy work quicker than traditional tools.

According to Riemann Computing, LonScript can complete certain complex calculations in seconds rather than minutes. The company has claimed performance of more than 1,400 times faster than Python in its benchmark testing (across certain tests), though further real-world results will depend on the task, hardware, and software setup.

The language is intended to help researchers and developers reduce the time and computing resources needed for AI training, simulations, and large-scale data work. Riemann Computing founder Andrew Kamal has described LonScript as a shift in how computers approach difficult problem-solving.

Together, the new model, the savings estimator, and LonScript reflect Riemann Computing's larger pitch: advanced technology should not be limited to organizations with enormous budgets. By focusing on compression, efficient compute, and faster scientific programming, the company aims to lower the cost of building and using modern software.

Riemann Computing is currently running an , where supporters can follow its progress and learn more about its work. As always, none of this is investment or financial advice. Please do your own due diligence and research.

Media Contact

Andrew M. K. Nassief, Riemann Computing Inc., 1 248-238-8245 1, [email protected],

SOURCE Riemann Computing Inc.