Riemann Computing launches prerelease AI model while fundraising on Wefunder

Riemann Computing has released a prerelease language model called “My Name is Riemann” as it raises capital on Wefunder and pushes a hybrid strategy that mixes open AI projects with proprietary enterprise tools. The company says the effort is designed to widen access to AI development for smaller teams, not just large tech firms.

Why it matters: - Riemann Computing is betting that AI development can become more accessible if infrastructure, training methods and model access are opened beyond large technology companies. - The company is pairing public-facing AI projects with planned commercial products, which could give smaller developers more ways to build, train and customize models. - The fundraising campaign on Wefunder adds a financing angle to the launch and puts the company’s long-term platform strategy in front of prospective supporters.

What happened: - Riemann Computing launched a prerelease language model called “My Name is Riemann” while currently fundraising on Wefunder. - Founder Andrew Magdy Kamal is leading the company’s push to expand who can build and adapt AI systems. - The company says its OpenPeer AI initiative is exploring decentralized infrastructure and alternative training approaches. - Developers can follow the company’s published projects at OpenPeerAI on Hugging Face.

The details: - The prerelease model has approximately 13.1 billion parameters. - Riemann Computing says the model was developed from published OpenPeerLLM weights through staged model growth. - The project uses the OpenPeer-AI-NTK-Trainer to continue model development in a more resource-conscious way. - The company says the initial training phase used about $3,000 in local hardware. - Riemann Computing says later stages used about $640 in cloud computing. - The company says those later stages ran on a DigitalOcean node with eight AMD MI325X GPUs for 21 node-hours, equal to 168 GPU-hours. - The company says those figures reflect hardware and cloud resources only, not labor, research, data preparation or prior model development. - Riemann Computing says the prerelease is an early demonstration, not a production-ready chatbot and not evidence of frontier-model-level performance. - The company intends the model to serve as a foundation for more training, experimentation and application development. - Riemann Computing was founded in 2018 and was formerly known as Stark Drones Corporation. - The company develops data compression, computing and decentralized infrastructure technologies. - Riemann Computing plans to commercialize and sell proprietary codecs, enterprise models and systems.

Between the lines: - The launch suggests Riemann Computing is trying to position itself as both an open ecosystem builder and a commercial infrastructure company. - The low reported training spend is part of the pitch, but the company also explicitly narrows what that figure does and does not include. - The prerelease framing signals an R&D milestone rather than a finished product claim, which helps set expectations for technical performance and market readiness.

What's next: - Riemann Computing says it will keep developing the model and broader OpenPeer AI ecosystem. - The company plans to use the prerelease as a base for future training and application work. - Prospective supporters can review the fundraising campaign at the Wefunder offering page, including offering terms and risk disclosures. - The company will also continue publishing project updates through its Hugging Face presence and social channels.

The bottom line: - Riemann Computing is using a prerelease model to showcase a broader thesis: smaller teams should be able to participate in AI development, while the company builds products it can later sell.